awesome-egocentric-atlas / awesome-egocentric-papers.csv
cy0307's picture
Sync July 30 catalog update
fde305b verified
Raw
History Blame Contribute Delete
351 kB
name,kind,released,venue,status,scope,year,paper,url,code,license,scale,tasks,modalities
ReViV,model,2026-07,ECCV 2026,open,egocentric,2026,https://arxiv.org/abs/2607.17790,https://reviv4d.github.io/,https://github.com/lvsean/reviv4d,Apache-2.0 code; non-commercial checkpoint terms,"Unified feed-forward monocular egocentric 4D reconstruction of RGB, camera trajectory, gaze, body, hands, and depth; evaluated on HoloAssist, HOT3D, ARCTIC, Aria Digital Twin, TACO, and pretrained on more than 600 hours",4d-reconstruction; camera-tracking; egocentric-3d-pose; gaze-estimation; depth-estimation,egocentric-video; camera-pose; gaze; body-pose; hand-pose; depth
EgoRecovery,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.19745,https://arxiv.org/abs/2607.19745,,,Co-training framework that aligns egocentric human failure-recovery demonstrations with a corrective-intent space and a small robot-data anchor; its collection protocol yields more than 10x as much valid recovery data per hour as robot teleoperation,failure-recovery; robot-learning; imitation-learning; human-to-robot-transfer,egocentric-video; human-demonstrations; robot-actions
Sidewalk Moments,dataset,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.20903,https://arxiv.org/abs/2607.20903,,,"61 first-person YouTube city-walk videos segmented into more than 50,000 ten-second clips with video, temporally averaged image, audio, and text representations for human-aligned urban-engagement analysis",urban-understanding; human-alignment; multimodal-representation; engagement-prediction,egocentric-video; audio; natural-language; image
CRAFT Wearable Creative AI,toolkit,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.21394,https://arxiv.org/abs/2607.21394,,,"Context-aware smart-glasses creative-AI probe studied through nine writer interviews, co-design with 16 writers and researchers, and 24 real-world sessions with eight writers",wearable-ai; context-aware-assistance; human-computer-interaction; creativity-support,smart-glasses; egocentric-video; natural-language
SHARE User-Centric AR SLAM,toolkit,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.23901,https://arxiv.org/abs/2607.23901,,,"User-prioritized edge SLAM for commercial head-mounted AR and a ground robot in shared workspaces; reports 13.22 ms average AR latency, a 43.3% reduction, while maintaining sub-2-centimeter tracking",ar-vr; slam; human-robot-collaboration; localization,ar-headset; egocentric-video; robot-video; camera-pose
EgoPlay,model,2026-07,SIGGRAPH Asia 2026,partial,egocentric,2026,https://arxiv.org/abs/2607.24560,https://egoplay2026.github.io/egoplay/,,,"Event-triggered egocentric video-to-video editing trained on 106K clip-prompt pairs, primarily from Ego4D, with positive, negative, and multi-event triggers plus event-aware evaluation",video-editing; event-detection; video-generation; streaming-video,egocentric-video; natural-language; generated-video
Data Pyramid for Embodied Manipulation,survey,2026-07,arXiv,open,adjacent,2026,https://arxiv.org/abs/2607.24744,https://jasper-aaa.github.io/embodied-data-pyramid/,https://github.com/worldbench/awesome-embodied-data-pyramid,apache-2.0,"Survey and living catalog organizing real-robot, UMI, human egocentric and ego-exo, simulation, and general video data into a five-level embodied-manipulation data pyramid",survey; data-curation; robot-manipulation; embodied-ai,
HiFi-UMI-2K,dataset,2026-07,arXiv,open,egocentric,2026,https://arxiv.org/abs/2607.25895,https://huggingface.co/datasets/simple-world-lab/HiFi-UMI-2K,,cc-by-4.0,"2,000 hours of public robot-free manipulation demonstrations from a high-fidelity UMI rig with head stereo-inertial SLAM, native inter-gripper pose, microsecond GPIO synchronization, and two roughly 200-degree cameras per hand; reports 3 mm local end-effector accuracy",robot-learning; imitation-learning; vision-language-action; world-action-model; human-to-robot-transfer,egocentric-video; stereo-video; wrist-video; camera-pose; gripper-pose; imu; robot-actions
EgoSafe-Bench,benchmark,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.26518,https://arxiv.org/abs/2607.26518,,,"12,000 safety-reasoning evaluation samples formed from 3,000 first-person mobile-captured clips and hierarchical QA chains spanning evidence anchoring, blind-spot deduction, intent inference, and causal consistency",safety-reasoning; video-qa; causal-reasoning; benchmark,egocentric-video; video-qa; natural-language
HumanCLAW-Bench,benchmark,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.27180,https://human-claw.github.io/,https://github.com/Human-CLAW/HumanCLAW,,"1,218 long-horizon egocentric find-navigate-interact episodes across 41 indoor scenes, evaluating nine VLMs through atomic full-body skill commands; the best reported success rate is 16.8%",embodied-reasoning; navigation; interaction; body-awareness; benchmark,egocentric-video; body-pose; natural-language; embodied-actions
ChildLens,dataset,2026-04,Behavior Research Methods 2026,request,egocentric,2026,https://doi.org/10.3758/s13428-026-02982-6,https://www.eva.mpg.de/comparative-cultural-psychology/technical-development/childlens/,https://github.com/neleSuffo/ChildLens,other,"109 hours of vest-mounted 140-degree egocentric video and audio from 62 children aged 3 to 5 in their homes, with five location classes and 14 exhaustively annotated activity classes",child-development; activity-recognition; temporal-localization; voice-type-classification,egocentric-video; audio; natural-language; annotations
HuMI,dataset,2026-02,arXiv,open,egocentric,2026,https://arxiv.org/abs/2602.06643,https://humanoid-manipulation-interface.github.io/,https://github.com/Richard-coder-Nai/HuMI,MIT code; CC-BY-4.0 datasets,"Portable robot-free whole-body demonstration interface and seven-dataset Hugging Face collection spanning five kneeling, squatting, tossing, walking, and bimanual tasks; reports 3x collection efficiency and 70% unseen-environment success",humanoid-manipulation; whole-body-control; imitation-learning; human-to-robot-transfer,egocentric-video; wrist-video; body-pose; hand-pose; robot-actions
ContactFlow,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.26579,https://arxiv.org/abs/2607.26579,,,"Embodiment-agnostic video action conditioning through trajectories of 3D actor-object contact points, trained with human and robot interactions and evaluated on DROID and real tabletop tasks",world-modeling; human-to-robot-transfer; contact-reasoning; robot-manipulation,human-video; robot-video; 3d-contact; generated-video
Robot-Factored World Models,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.22535,https://bjkim95.github.io/rofacto/,https://github.com/bjkim95/rofacto,,"Action-conditioned world model that factors robot control and appearance into nominal trajectories and rendered robot geometry, evaluated on DROID and egocentric RoboCasa-GR1 with zero-shot embodiments and human-demo retargeting",world-modeling; robot-manipulation; cross-embodiment-transfer; video-generation,robot-video; depth; robot-actions; generated-video
EgoExoMoCap,model,2026-07,ECCV 2026,watch,egocentric,2026,https://arxiv.org/abs/2607.15868,https://arxiv.org/abs/2607.15868,,,Distributed human-motion capture in which two or more people wear smart glasses; fuses head and optional wrist tracking with DINOv3 image features and is evaluated on two in-the-wild ego-exo datasets,egocentric-3d-pose; human-motion; ego-exo; human-pose-estimation,hmd-video; exocentric-video; head-pose; wrist-trajectory; body-pose
Exo2EgoPose,model,2026-07,ACM MM 2026,watch,egocentric,2026,https://arxiv.org/abs/2607.15890,https://arxiv.org/abs/2607.15890,,,"Vision-language-guided egocentric 3D hand-pose forecasting that reconstructs exocentric demonstrations to guide ego-view prediction, evaluated on AssemblyHands, Ego-Exo4D, EgoMe-pose, and CALVIN transfer",3d-hand-trajectory-forecasting; hand-forecasting; ego-exo; cross-embodiment-transfer,egocentric-video; exocentric-video; 3d-hand-pose; language-instructions
Let the Body Follow,toolkit,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.16095,https://arxiv.org/abs/2607.16095,,,Coupled egocentric whole-body teleoperation for TIAGo in which head and arm motion automatically coordinate torso and mobile-base control; evaluated in a home-care-inspired user study,digital-teleoperation; whole-body-control; human-robot-interaction; mobile-manipulation,hmd-video; head-pose; hand-pose; robot-actions
VideoTreeSearch,model,2026-07,arXiv,open,egocentric,2026,https://arxiv.org/abs/2607.16189,https://github.com/CeeZh/VTS,https://github.com/CeeZh/VTS,MIT,"Self-correcting temporal-tree search agent for grounded long-video QA, with SFT and RL code, an 8B checkpoint, and released trajectories/evaluation data; gains 7.4 T-F1 on Haystack-Ego4D",long-video-qa; temporal-grounding; memory-reasoning; agentic-reasoning,long-video; video-qa; natural-language
EgoPoseVR,dataset,2026-02,IEEE VR 2026 / IEEE TVCG,open,egocentric,2026,https://arxiv.org/abs/2602.05590,https://aplusx.github.io/EgoPoseVRWeb/,,cc-by-4.0,"18,235 synthetic 100-frame RGB-D/HMD motion clips across seven VR scenes, derived from 2,344 AMASS sequences and split 14,702/1,827/1,706 for train/validation/test; 388 GB on Hugging Face",egocentric-3d-pose; human-pose-estimation; ar-vr; human-motion,synthetic-egocentric; rgb-d; head-pose; body-pose
EgoObjects-Camera,dataset,2026-07,Hugging Face,partial,egocentric,2026,https://arxiv.org/abs/2510.08673,https://huggingface.co/datasets/KangLiao/EgoObjects-Camera,https://github.com/KangLiao929/Puffin,other,"Public 508 MB annotation release with predicted roll, pitch, vertical field of view, radial distortion, and parse validity for 241,554 EgoObjects images across 49 shards",calibration; egocentric-vision; spatial-reasoning; data-augmentation,egocentric-images; camera-orientation; camera-intrinsics; annotations
"Think at 5 Hz, Act at 20 Hz",model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.15621,https://arxiv.org/abs/2607.15621,,,"Asynchronous driving VLA with a frozen 7B slow reasoner and fast action expert, producing fresh control every 50 ms and raising CARLA LangAuto-Short route completion from 37.0 to 94.0",vla; driving; efficient-inference; robot-control,vehicle-camera-video; language-instructions; trajectories
IMBench,benchmark,2026-07,RSS 2026 SemRob Workshop,open,adjacent,2026,https://arxiv.org/abs/2607.15641,https://imbench.org/,,apache-2.0,"Intuitive-manipulation benchmark spanning 35 tasks, seven categories, and 14K filtered trajectories, jointly testing perception, physical reasoning, action generation, and iterative execution",robot-manipulation; physical-world-understanding; planning; benchmark,robot-camera-video; language-instructions; trajectories; robot-actions
AC-VLA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.15714,https://arxiv.org/abs/2607.15714,,,"Plug-and-play compositional VLA training with instruction decomposition, trajectory alignment, and state-conditioned wrist-view masking; reports about 28% absolute improvement on LIBERO compositional OOD tasks",vla; compositional-generalization; robot-manipulation; domain-generalization,wrist-camera; robot-camera-video; language-instructions; robot-actions
SkillNav,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.15758,https://arxiv.org/abs/2607.15758,,,"Training-free object-goal navigation framework that writes composable behavioral skills into a curiosity-value map, reaching SPL 25.5 on MP3D and 43.2 on HM3D v0.2",navigation; memory; semantic-mapping; zero-shot,robot-camera-video; spatial-memory; trajectories
Orbis 2,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.15898,https://lmb-freiburg.github.io/orbis2.github.io/,,,Hierarchical driving world model separating long-horizon coarse structure from detailed generation and combining diffusion-forcing pretraining with teacher-forcing fine-tuning,world-modeling; driving; video-generation; representation-learning,vehicle-camera-video; generated-video; trajectories
Data and Learning Where it Matters,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.15982,https://arxiv.org/abs/2607.15982,,,"Targeted autonomous data collection and offline RL for contact-critical task segments, combined with conventional free-space planning; averages 96% success across four real tasks using 2 to 2.5 hours of data",contact-rich-manipulation; data-collection; reinforcement-learning; planning,robot-camera-video; robot-actions; trajectories
MotionForesight,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.16192,https://motionforesight.github.io/,,,"Repurposes frozen video-prediction and 3D-tracking priors to forecast future object-centered 3D scene flow from short monocular human-object video context, trained on 40K videos",future-prediction; object-motion; world-modeling; contact-understanding,monocular-video; trajectories; 3d
Ego Scene Augmentation (ESA),model,2026-07,arXiv,partial,egocentric,2026,https://arxiv.org/abs/2607.14497,https://github.com/Chikit-WONG/spatialGraph,https://github.com/Chikit-WONG/spatialGraph,,"Ego-element Graph framework that combines visual-foundation-model signals into an explicit egocentric spatial representation, improving EgoTextVQA by 8.14% indoors and 8.72% outdoors",egocentric-vqa; spatial-reasoning; scene-understanding; representation-learning,egocentric-video; scene-graphs; spatial-features; natural-language
VIABench,benchmark,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.14660,https://github.com/MCG-NJU/VIABench,,,"First-person video benchmark recorded or shared by blind and visually impaired people, evaluating proactive navigation reminders, visual question answering, and vision-guided interaction in online and offline settings",blind-assistance; proactive-assistance; assistive-videoqa; vision-language; benchmark,egocentric-video; video-qa; natural-language
Wearable AI Dataset,benchmark,2026-05,ECCV 2026 Workshop,request,egocentric,2026,https://arxiv.org/abs/2606.04970,https://huggingface.co/datasets/facebook/wearable-ai,,cc-by-nc-4.0,"2,100 head-mounted everyday-activity videos (about 317 GB), split evenly across EgoConv conversational QA, EgoLongQA long-video multiple-choice QA, and EgoProactive streaming-assistant evaluation, with 700 samples per task and a bundled starter kit",long-video-qa; multimodal-conversation; proactive-assistance; wearable-assistants; benchmark,egocentric-video; video-qa; dialogue; natural-language
HUI360,dataset,2025-09,FG 2026,partial,egocentric,2025,https://doi.org/10.1109/FG67764.2026.11556969,https://hucebot.github.io/hui360/,https://github.com/hucebot/HUI360-Baselines,,"99 in-the-wild 360-degree egocentric HRI recordings (70 INRIA Shelfy and 29 Cornell SSUP-HRI), with open per-track skeleton, mask, box, and interaction annotations plus baselines for anticipation",human-robot-interaction; interaction-understanding; human-pose-estimation; anticipation; benchmark,first-person-360-video; skeleton; masks; annotations
EgoDyn-Bench,benchmark,2026-04,ECCV 2026,open,adjacent,2026,https://arxiv.org/abs/2604.22851,https://huggingface.co/datasets/TUM-AVS/EgoDyn-Bench-ECCV2026,https://github.com/TUM-AVS/EgoDyn-Bench,cc-by-nc-sa-4.0,"1,000 three-second clips (500 nuScenes and 500 CARLA/Cosmos-transferred), 14 question types and about 14,000 oracle-labelled QA pairs, per-clip dynamics arrays, and a 49-model reference leaderboard",spatial-reasoning; physical-world-understanding; video-qa; benchmark; driving,vehicle-camera-video; video-qa; trajectories; annotations
VTM-Nav,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.14514,https://arxiv.org/abs/2607.14514,,,"Training-free cross-episode object-goal navigation with persistent room- and object-level visual-topological memory, coarse-to-fine retrieval, and a conservative execution guard, evaluated on HM3D v0.1/v0.2 and MP3D",navigation; memory; semantic-mapping; long-horizon-reasoning,robot-camera-video; spatial-memory; scene-graphs; trajectories
SafeRelBench,benchmark,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.14543,https://arxiv.org/abs/2607.14543,,,"507 executable household-agent evaluations, including 248 spatial-relation safety cases and 259 controls, testing whether VLM-driven agents satisfy support, containment, and proximity constraints before risk-prone actions",safety; spatial-reasoning; benchmark; embodied-reasoning,robot-camera-video; language-instructions; annotations
SoftNav,model,2026-07,IROS 2026,watch,adjacent,2026,https://arxiv.org/abs/2607.14586,https://arxiv.org/abs/2607.14586,,,"Injects one continuous 3D soft token per detected object or frontier into a frozen VLM using about 1,200 samples and 17M trainable parameters, with transfer across HM3D-OVON, GOAT-Bench, SG3D, and a real robot",navigation; 3d-scene-understanding; representation-learning; domain-generalization,robot-camera-video; point-clouds; spatial-features; language-instructions
Representation-Aligned Tactile Grounding,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.14609,https://arxiv.org/abs/2607.14609,,,"Latent Tactile Predictor grounded at intermediate action-expert features, selected through linear probing to align VLA representations with future contact consequences without predicting noisy raw tactile signals",tactile-learning; contact-rich-manipulation; vla; robot-learning,robot-camera-video; tactile; robot-actions; latent-features
Action QFormer,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.14635,https://arxiv.org/abs/2607.14635,,,"Instruction-conditioned query interface that reorganizes inherited multimodal features before action generation, raising zero-shot sim-to-real navigation success from 18.8% to 56.3% and action correctness from 22.5% to 75.5%",vla; navigation; representation-learning; domain-generalization,robot-camera-video; language-instructions; robot-actions; latent-features
Reflex,toolkit,2026-07,ICML 2026,watch,adjacent,2026,https://arxiv.org/abs/2607.14695,https://arxiv.org/abs/2607.14695,,,"Streaming flow-matching VLA inference with static/sliding/dynamic attention regions, AdaRMSNorm, asynchronous visual encoding, and operator fusion, reporting 2.58x speedup, stable 50 Hz control, and up to 54% lower reaction latency",vla; efficient-inference; robot-deployment; robot-control,robot-camera-video; robot-actions; latent-features
FoMoVLA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.14739,https://liauto-research.github.io/FoMoVLA/,,,"Couples compact future-feature foresight with sparse 2D point trajectories through future-conditioned cross-attention, supplying both target-state and motion-path supervision for continuous VLA policies",vla; future-prediction; tracking; robot-learning,robot-camera-video; robot-actions; latent-features; trajectories
LifelongVLA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.14852,https://arxiv.org/abs/2607.14852,,,"Dual-timescale LoRA gating separates short-term plasticity from long-term consolidation, paired with cache-efficient stochastic replay for sequentially expanding manipulation skills on an xArm robot",vla; continual-learning; robot-learning; manipulation,robot-camera-video; robot-actions; language-instructions
Steering Robustness into World Action Models,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.14943,https://arxiv.org/abs/2607.14943,,,"Mechanistic analysis of robustness-relevant WAM activations plus contrastive steering directions and a reduced-order World-Action LQR controller, improving robustness on Cosmos-Policy and DiT4DiT",world-action-models; robustness; robot-control; evaluation,robot-camera-video; robot-actions; latent-features
AeroAct,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.14997,https://arxiv.org/abs/2607.14997,,,"First reported real-world aerial WAM, adapting a video diffusion Transformer to predict local quadrotor trajectory-action chunks from egocentric visual history, proprioception, and language without decoding future video at deployment",world-action-models; navigation; vla; robot-control,uav-video; proprioception; language-instructions; robot-actions
DriftWorld,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.15065,https://susie-lu.github.io/driftworld/,,,"Single-forward-pass action-conditioned world model generating robot future frames at more than 30 fps, reported 17x faster on average than diffusion baselines and reaching up to 0.99 correlation for offline policy ranking",world-modeling; planning; robot-control; evaluation,robot-camera-video; robot-actions; generated-video
BadWAM,benchmark,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.15207,https://arxiv.org/abs/2607.15207,,,World-Action Drift Attack framework testing action-only and imagination-preserving perturbations; the reported action-only attack lowers one WAM from 96.5% to 43.1% closed-loop success,world-action-models; robustness; safety; benchmark,robot-camera-video; robot-actions; generated-video
RoboTTT,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.15275,https://research.nvidia.com/labs/gear/robottt/,,,"Test-time-training robot policy scaling visuomotor context to 8K timesteps without growing inference latency, enabling one-shot imitation from a human video, online improvement, perturbation recovery, and a five-minute ten-stage assembly",vla; long-horizon-reasoning; imitation-learning; test-time-scaling; manipulation,robot-camera-video; human-view-video; robot-actions; language-instructions
EgoProceVQA,benchmark,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.13792,https://z1oong.github.io/EgoProceVQA/,,,"1,272 source clips and 3,600 key-step question-answer pairs across 31 procedural tasks and four scenarios, covering six question types plus the EgoProceGen data platform and self-skill-exploring EgoProceAgent",egocentric-vqa; procedural-learning; key-step-localization; long-horizon-reasoning; procedural-assistance,egocentric-video; video-qa; annotations; natural-language
EgoHTR,dataset,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.13472,https://egohtr.github.io/,,,"55 scene-aligned 4D human-motion sequences across seven challenging terrains, totaling 1.37 hours and about 150K frames from eight participants, with 36 multiview sequences and 0.7 hours of motion-capture ground truth",4d-reconstruction; human-motion; humanoid-control; scene-aware-3d-human-pose; robot-learning,egocentric-video; multiview-video; 3d-human-motion; scene-scans; mocap; camera-pose
HRIBench (Interaction-Centric VLA),benchmark,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.13056,https://arxiv.org/abs/2607.13056,,,"13 executable role-conditioned collaboration tasks and 650+ episodes across Instructor, Collaborator, and Intruder roles, with synchronization, responsiveness, protocol, and safety metrics; simulation adaptation raises physical GR00T N1.5 success from 0.10 to 0.43",human-robot-interaction; vla; benchmark; robot-learning; safety,robot-camera-video; robot-actions; language-instructions; human-demonstrations
JITOMA / JITOMA-Bench,benchmark,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.13245,https://arxiv.org/abs/2607.13245,,,"Just-in-time scene-graph and memory activation for long-horizon robots, using dormant anchors and task-triggered local captioning to reduce active graph size and latency while sustaining multi-task reasoning",semantic-mapping; memory; long-horizon-reasoning; robot-learning; navigation,robot-camera-video; scene-graphs; natural-language; spatial-memory
REAL / REAL-Bench,toolkit,2026-07,ECCV 2026,open,adjacent,2026,https://arxiv.org/abs/2607.13653,https://github.com/InternRobotics/REAL,https://github.com/InternRobotics/REAL,MIT,"Open-world mobile-manipulation framework and 241-task REAL-Bench spanning exploration, distractors, articulated manipulation, and user disambiguation; reports 56.9% interactive success and 78.3% over 60 physical dual-arm mobile-robot episodes",mobile-manipulation; tool-using-agents; active-perception; human-robot-interaction; benchmark,robot-camera-video; robot-actions; language-instructions; simulation
VSI-Super-Wild,benchmark,2026-07,ECCV 2026,partial,adjacent,2026,https://arxiv.org/abs/2607.13681,https://vsi-super-wild.github.io/,https://github.com/THUSI-Lab/VSI-Super-Wild,MIT,"6,980 human-verified spatial QA pairs over 442 continuous panoramic in-the-wild videos totaling 284.52 hours across eight scene categories, including recordings longer than four hours and four observer-object-environment task families",long-video-understanding; spatial-reasoning; memory; video-qa; benchmark,360-video; long-video; video-qa; annotations
Anchor-Align,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.13429,https://anchoralignvla.github.io/,,,"VLA finetuning with frozen vision-language representation anchoring and language-action alignment, improving physical xArm7 success from 28% to 54% and from 37% to 60% across two VLA architectures",vla; domain-generalization; robot-learning; manipulation; language-grounding,robot-camera-video; robot-actions; language-instructions; latent-features
GPUSimBench,benchmark,2026-07,IROS 2026,watch,adjacent,2026,https://arxiv.org/abs/2607.13059,https://arxiv.org/abs/2607.13059,,,"Benchmark for GPU robot simulators covering inclined-plane real-to-sim physical consistency, throughput and memory scaling, plus run-to-run and inter-environment nondeterminism in Isaac Lab and Genesis",simulation; benchmark; systems-optimization; robot-learning,simulation; evaluation-logs; physics; robot-actions
WANDA / Worlds in One Demo,dataset,2026-07,arXiv,open,adjacent,2026,https://arxiv.org/abs/2607.13154,https://wanda.lecar-lab.org/,,cc-by-4.0,"Synthetic-data engine expanding one real demonstration per task into 16,922 LeRobot episodes and about 26.8M timesteps (roughly 248 hours at 30 fps) across five long-horizon mobile-manipulation tasks and diverse generated worlds",data-generation; mobile-manipulation; world-modeling; imitation-learning; cross-embodiment-transfer,robot-camera-video; rgb-d; 3d-gaussian-splatting; robot-actions; synthetic-video
Reverse to Advance,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.13455,https://arxiv.org/abs/2607.13455,,,"Learns difficult manipulation policies by collecting easier reverse-direction rollouts, temporally inverting and filtering them with kinematic and critic priors, then iterating hard-task policy training with automated reset cycles",imitation-learning; data-collection; manipulation; robot-learning,robot-camera-video; robot-actions; offline-demonstrations
JOP-VLN,model,2026-07,IROS 2026,watch,adjacent,2026,https://arxiv.org/abs/2607.13461,https://qingrongh.github.io/JOP-VLN,,,"Three-stage VLN training that combines imitation learning, DAgger recovery trajectories, and on-policy reinforcement learning, reaching 69.9% success on R2R and 68.0% on RxR",navigation; reinforcement-learning; imitation-learning; vla; exploration,robot-camera-video; language-instructions; trajectories
UESF-Bench / SeekFollow-VLA,benchmark,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.13621,https://arxiv.org/abs/2607.13621,,,"Unified language-guided human seeking and persistent following benchmark requiring semantic exploration, delayed identity grounding, phase switching, and recovery in single- and multi-person environments",navigation; human-robot-interaction; vla; tracking; person-identification,robot-camera-video; language-instructions; trajectories; annotations
GigaWorld-Policy-0.5,model,2026-07,arXiv,open,adjacent,2026,https://arxiv.org/abs/2607.13960,https://open-gigaai.github.io/giga-world-policy/,https://github.com/open-gigaai/giga-world-policy,apache-2.0,"Action-centered WAM combining action-conditioned world modeling, mixed WAM training, Mixture-of-Transformers experts, and agentic AutoResearch; reports 85 ms local inference on an RTX 4090 without decoding future video at deployment",world-action-models; vla; efficient-inference; robot-control; manipulation,robot-camera-video; robot-actions; generated-video; latent-features
Industrial Dexterity Benchmark,benchmark,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.14021,https://arxiv.org/abs/2607.14021,,,"Hardware benchmark boards for datacenter cabling, automotive harnesses, and gearbox assembly, plus DAG-ROS and multimodal AG-iDP3; the strongest cable configuration reaches 78% versus 36% for single-camera RGB with about 100 demonstrations per phase",dexterous-manipulation; benchmark; imitation-learning; contact-rich-manipulation; robot-learning,robot-camera-video; point-clouds; force; robot-actions; proprioception
PhysClaw-0,toolkit,2026-07,arXiv,open,adjacent,2026,https://arxiv.org/abs/2607.14047,https://open-gigaai.github.io/PhysClaw/,https://github.com/open-gigaai/PhysClaw,apache-2.0,"Autonomous collect-verify-reset system that converts operator language corrections into reusable structured memory, reducing human working time to 16% and raising average single-attempt success from 12.5% to 47.5%",data-collection; robot-learning; manipulation; expert-feedback; tool-using-agents,robot-camera-video; robot-actions; language-instructions; logs
DenseReward,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.13033,https://arxiv.org/abs/2607.13033,,,"Dense frame-level vision-language reward model trained with automatically synthesized physical failures such as collisions, missed grasps, drops, and recovery behaviors, evaluated for simulated and real manipulation plus MPC and RL guidance",robot-learning; reward-modeling; reinforcement-learning; manipulation; failure-understanding,robot-camera-video; language-instructions; synthetic-failures; dense-rewards
FlowWAM,model,2026-07,arXiv,open,adjacent,2026,https://arxiv.org/abs/2607.13017,https://flow-wam.github.io/,https://github.com/YixiangChen515/FlowWAM,Apache-2.0,Dual-stream world-action model using optical flow as a shared video-native action representation; reports 92.94% RoboTwin Clean and 92.14% Random success plus a 63.71 WorldArena EWMScore,world-action-models; vla; world-modeling; robot-learning; manipulation,robot-camera-video; optical-flow; robot-actions; generated-video
ChunkFlow,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.12992,https://arxiv.org/abs/2607.12992,,,"Continuity-consistent chunked-policy framework with frozen, editable, and future zones, overlap blending, seam and derivative losses, history corruption, scheduled sampling, and AWAC fine-tuning",vla; robot-learning; action-chunking; manipulation; policy-stability,robot-camera-video; robot-actions; language-instructions
ExToken,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.12931,https://arxiv.org/abs/2607.12931,,,"Structured VLA reinforcement fine-tuning that derives discrete behavioral priors from offline demonstrations, conditions rollout exploration on tokens, and learns a state-conditioned token selector for deterministic deployment",vla; reinforcement-learning; robot-learning; exploration; manipulation,robot-camera-video; robot-actions; language-instructions; offline-demonstrations
Hy-Embodied-VLM-1.0,model,2026-07,arXiv,open,adjacent,2026,https://arxiv.org/abs/2607.12894,https://github.com/Tencent-Hunyuan/HY-Embodied,https://github.com/Tencent-Hunyuan/HY-Embodied,Apache-2.0,"Approximately 30B-parameter embodied MoE VLM activating about 3B parameters per token; evaluated on 38 physical-world benchmarks and ranked first among compared similarly sized models on 19, with an 8.4% average gain over HY-Embodied-0.5",embodied-ai; spatial-reasoning; navigation; long-horizon-reasoning; physical-world-understanding,images; video; natural-language; embodied-observations
Jetson-PI,toolkit,2026-07,arXiv,open,adjacent,2026,https://arxiv.org/abs/2607.12659,https://github.com/PKU-SEC-Lab/Jetson-PI,https://github.com/PKU-SEC-Lab/Jetson-PI,Apache-2.0,"Foresight-aligned asynchronous VLA inference for Jetson Orin, reporting 8.66x higher control frequency than naive PyTorch, 5.41x over vla.cpp, and a 14.8% average success-rate gain over VLASH on LIBERO",vla; edge-inference; robot-control; manipulation; systems-optimization,robot-camera-video; robot-actions; latent-features
TrustVLA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.12571,https://arxiv.org/abs/2607.12571,,,"Retraining-free defense that detects abnormal VLA evidence evolution, localizes compact visual-trigger regions by counterfactual score drop, and recovers observations through localized inpainting",vla; safety; backdoor-defense; robot-learning; manipulation,robot-camera-video; robot-actions; visual-triggers
Self-in-Space / SIS-Bench,benchmark,2026-07,ACM MM 2026,open,adjacent,2026,https://arxiv.org/abs/2607.12477,https://choucisan.github.io/publications/self-in-space/,https://github.com/IntelliSensing/Self-in-Space,Apache-2.0,"4,856 expert-verified QA pairs across 13 self-awareness and spatial-cognition tasks from 1,646 real-world UAV videos, plus 54,298 public SIS-Motion training examples and an open motion-aware adapter",embodied-ai; spatial-reasoning; self-awareness; video-qa; navigation,uav-video; optical-flow; natural-language; annotations
VistaVLA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.12356,https://arxiv.org/abs/2607.12356,,,"Grounds geometry- and semantics-aware 3D Gaussian primitives into compact VLA context tokens; reports 99% token reduction, +22.8% success over seven real tasks, and +30.0% over VLA-Adapter on OOD tasks",vla; 3d-spatial-reasoning; robot-learning; manipulation; domain-generalization,multiview-video; 3d-gaussians; robot-actions; language-instructions
Efficient VLA Inference via Temporal Redundancy Reduction,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.12287,https://arxiv.org/abs/2607.12287,,,"System-level VLA acceleration that updates only dynamic visual tokens and compresses diffusion action sampling to two steps, reporting more than 2x speedup and up to 98% success across LIBERO, RoboTwin, and real robots",vla; efficient-inference; robot-control; manipulation; systems-optimization,robot-camera-video; robot-actions; visual-tokens
Vinci2 / EgoServe,benchmark,2026-07,ECCV 2026,open,egocentric,2026,https://arxiv.org/abs/2607.11523,https://sitonggong.github.io/EgoServe-page/,https://github.com/SitongGong/EgoMemo,cc-by-4.0,"3,000+ proactive-service instances across 11 service categories and four temporal memory horizons, with the training-free EgoMemo agent combining multi-scale summaries, a semantic graph, and visual retrieval archives",proactive-assistance; memory; long-context; video-qa; streaming-assistant,egocentric-video; annotations; natural-language; memory
LightMem-Ego,toolkit,2026-07,arXiv,open,egocentric,2026,https://arxiv.org/abs/2607.11487,https://github.com/zjunlp/LightMem-Ego,https://github.com/zjunlp/LightMem-Ego,MIT,"Streaming multimodal memory system that aligns first-person visual and audio capture into current, short-term, and long-term memory for object finding, conversation recall, life summaries, routine discovery, and evidence-grounded answers",memory; long-horizon-memory; wearable-assistants; personalized-qa; visual-assistance,egocentric-video; audio; speech; smart-glasses; memory
EgoSteer,model,2026-06,arXiv,open,egocentric,2026,https://arxiv.org/abs/2607.09701,https://egosteer.github.io/,https://github.com/egosteer/egosteer,Apache-2.0,"Full-stack dexterous VLA system using EgoSmith to curate 9.6K hours of egocentric human video, an open robot stack for teleoperation and correction, and 3B base and RealMan checkpoints; evaluated on 40+ tasks with 75%+ few-shot success on long-horizon tasks",vla; dexterous-manipulation; human-demonstrations; cross-embodiment-transfer; world-action-models,egocentric-video; hand-motion; robot-camera-video; robot-actions; language-instructions
InterPet4D,dataset,2026-06,Hugging Face,open,egocentric,2026,https://arxiv.org/abs/2607.10287,https://huggingface.co/datasets/ohicarip/interpet4d,,cc-by-nc-4.0,"113 human-dog interaction sessions with 6.8M synchronized multi-view and egocentric frames from 13 dogs across 11 breeds and 23 people; public v1 provides 227 headset clips with audio and aligned SMPL, MANO, pet skeleton, SMAL, and MERT features",3d-4d-understanding; human-pose-estimation; motion-generation; animal-activity-recognition; interaction-understanding,egocentric-video; multiview-video; audio; 3d-human-motion; hand-motion; skeleton; mesh
REGRIND,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.11874,https://yunhaifeng.com/REGRIND,,,Retargeting-guided residual reinforcement-learning recipe that learns contact-rich dexterous policies from one human demonstration and transfers zero-shot to LEAP and WUJI hands for scissors and screwdriver tasks,dexterous-manipulation; retargeting; reinforcement-learning; contact-rich-manipulation; human-demonstrations,human-demonstrations; hand-object-motion; object-pose; robot-actions; simulation
TeleDexter,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.11481,https://bigai-dex.github.io/blog/teledexter/,,,Hand-object co-tracking controller trained with single-stage RL and transferred zero-shot across two dexterous hands; reports 75.2% average success over seven in-hand and tool-use tasks and uses 50 demonstrations per task for behavior cloning,digital-teleoperation; dexterous-manipulation; human-demonstrations; imitation-learning,human-demonstrations; hand-motion; object-pose; robot-actions
WALA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.11397,https://liujiahao2077.github.io/WALA.github.io,,,"Learns executable latent actions jointly from action-labeled demonstrations and action-free human or robot videos using DINOv3 feature deltas, dense depth, and a trainable latent world model; reports 75.2% RoboCasa success",latent-action-learning; world-action-models; vla; robot-learning; human-demonstrations,robot-camera-video; human-demonstrations; dense-features; depth; latent-actions; robot-actions
Robot-Centric Pointmaps,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.11498,https://davian-robotics.github.io/pointmap/,,,"Encodes per-pixel XYZ in the robot frame while preserving a dense image grid; improves pi0.5 by 7.6 points and SmolVLA by 4.2 over 24 RoboCasa tasks, with a +11.7-point gain at an unseen viewpoint in four real tasks using 180 demonstrations",vla; spatial-reasoning; robot-learning; manipulation; domain-generalization,robot-camera-video; point-clouds; depth; robot-actions
Lumo-2,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.11270,https://arxiv.org/abs/2607.11270,,,"Latent world-action model with staged alignment of actions to latent dynamics, vision, and language, evaluated for scaling and out-of-distribution generalization on long-horizon and dexterous real-world manipulation",world-action-models; vla; robot-learning; long-horizon-reasoning; dexterous-manipulation,robot-camera-video; latent-actions; robot-actions; language-instructions
Xiaomi-Robotics-U0,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.11643,https://arxiv.org/abs/2607.11643,,,"38B unified embodied-synthesis model spanning multi-view scene generation, embodiment transfer, editing, and video generation; reports improving pi0.5 out-of-distribution manipulation success from 36.9% to 63.2%",world-modeling; data-generation; robot-learning; video-generation; cross-embodiment-transfer,generated-video; multiview-video; robot-camera-video; robot-actions
World Action Models to Embodied Brains Roadmap,survey,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.11689,https://arxiv.org/abs/2607.11689,,,"Roadmap organizing WAM gaps across model roles and representations, objectives and standardization, and system composition, then proposing embodied brains, physical harnesses, shared contracts, and closed-loop post-training",world-action-models; embodied-ai; survey; robot-learning,survey; embodied-models
SLVMBench,benchmark,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.11312,https://arxiv.org/abs/2607.11312,,,"2,261 human-authored questions over 1,220 videos in 11 domains, embedding tutorials inside 2-3-hour distractor streams to test whether video LLMs retain procedural knowledge and apply it to an ongoing target video",procedure-learning; long-video-qa; memory; evaluation,long-video; video-qa; temporal-localization
E-VQA / ST-Evidence,benchmark,2026-07,ECCV 2026,open,adjacent,2026,https://arxiv.org/abs/2607.11862,https://github.com/SalesforceAIResearch/EVQA,https://github.com/SalesforceAIResearch/EVQA,cc-by-nc-4.0,"Human-verified benchmark for answers plus temporal segments and tracked object masklets, accompanied by 160K ST-Evidence-Instruct examples and released 3B and 7B grounded video-language models",video-qa; visual-grounding; segmentation; evidence-consistency; evaluation,video; video-qa; object-masks; temporal-localization; annotations
DA-Nav / ReDA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.11638,https://arxiv.org/abs/2607.11638,,,"Direction-aware city-scale VLN framework grounding commercial navigation instructions on the egocentric image plane, with the ReDA recovery dataset; reports 56.16% success in unseen CARLA cities and kilometer-scale quadruped and humanoid trials",navigation; spatial-reasoning; embodied-ai; domain-generalization,robot-camera-video; language-instructions; trajectories; maps
AdvNav,model,2026-07,ACM MM 2026,watch,adjacent,2026,https://arxiv.org/abs/2607.11063,https://arxiv.org/abs/2607.11063,,,Behavior-guided black-box attack on first-person VLN observations using trajectory and action feedback plus adaptive and genetic optimization; reports 49.70%-87.30% attack success against HAMT and MapGPT on R2R,navigation; robustness; evaluation; embodied-ai,robot-camera-video; trajectories; language-instructions
TSR-Ego,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.09169,https://arxiv.org/abs/2607.09169,,,"Online egocentric 3D human-pose framework using causal temporal stereo features, temporal and joint self-attention, and fisheye deformable stereo cross-attention to recover occluded or truncated body joints",egocentric-3d-pose; human-pose-estimation; pose-estimation; motion-estimation,egocentric-video; stereo-video; fisheye-video; 3d-human-pose
AgenticFocus,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.08857,https://arxiv.org/abs/2607.08857,,,"Mixed-reality synthesis pipeline that restores occluded object geometry and full-hand motion from ordinary first-person human video, then retargets and composites the interaction into synchronized humanoid visual observations, actions, and states",human-demonstrations; robot-learning; cross-embodiment-transfer; retargeting; dexterous-manipulation,egocentric-video; hand-motion; object-meshes; robot-actions
TactiDex,benchmark,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.09190,https://tactidex.github.io/,,,"Real-world human-demonstration benchmark synchronizing whole-hand tactile pressure, multi-granularity hand kinematics, object 6D states, language descriptions, and task phases, with TactiSkill transfer across single-hand and bimanual tasks",dexterous-manipulation; human-demonstrations; cross-embodiment-transfer; contact-understanding; bimanual-manipulation,tactile; hand-motion; object-pose; language-instructions
DemoBridge,toolkit,2026-07,RSS 2026 RoboData Workshop,open,adjacent,2026,https://arxiv.org/abs/2607.09519,https://gitlab.kuleuven.be/u0123974/demo-bridge/,https://gitlab.kuleuven.be/u0123974/demo-bridge/,Apache-2.0,"Simulation-in-the-loop toolkit converting a single-view RGB stereo human-hand demonstration into a collision-aware, physics-validated robot-arm trajectory; evaluated on three real-demonstration tasks and a synthetic planner benchmark",human-demonstrations; retargeting; robot-learning; simulation; trajectory-synthesis,video; stereo-video; hand-motion; robot-actions; simulation
PhysV2A,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.09365,https://arxiv.org/abs/2607.09365,,,"Reachability-gated feasibility-completion framework converting video-derived 6D object motion into robot-executable trajectories through grasp-conditioned hypotheses, kinematic checks, and VLM-assisted semantic-mask relaxation; evaluated on four real tabletop tasks",human-demonstrations; robot-learning; cross-embodiment-transfer; manipulation; trajectory-synthesis,rgbd; object-pose; robot-actions; semantic-masks
OpenGlass Visual Assistance,toolkit,2026-07,ACL 2026 System Demonstrations,partial,egocentric,2026,https://arxiv.org/abs/2607.03213,https://github.com/OpenSQZ/OpenGlass,https://github.com/OpenSQZ/OpenGlass,,"Local-first visual-assistance reference system splitting ESP32 glasses sensing from nearby-device MLLM inference; reports 993 ms median user-to-audio latency for resized images and releases source, hardware instructions, prompts, evaluation data, and logs",visual-assistance; accessibility; edge-ai; on-device-perception; wearable-interaction,egocentric-video; smart-glasses; audio; speech; evaluation-logs
Worldscape-MoE,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.03964,https://worldscape-moe.com/,https://github.com/EmbodiedCity/Worldscape-MoE.code,,"Mixture-of-experts diffusion world model unifying camera trajectories, robot actions, and hand-joint action maps across locomotion, manipulation, and egocentric hand-control experiments",world-modeling; world-action-models; video-generation; hand-object-interaction; robot-learning,egocentric-video; generated-video; hand-motion; robot-actions; camera-trajectories
Trustworthy Visual Predicates,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.08121,https://arxiv.org/abs/2606.08121,,,"Predicate-level reliability framework evaluating contact, support, containment, motion coupling, grasp, release, and active-hand evidence under blur, occlusion, illumination change, low resolution, frame dropping, and detection noise",hand-object-interaction; manipulation-understanding; robustness; evaluation; neuro-symbolic-reasoning,egocentric-video; bimanual-video; hand-object-contact; predicate-graphs
Immersive Social Interaction with VR and LLM-Assisted Humanoids,toolkit,2024-11,Humanoids 2024 Workshop,watch,egocentric,2024,https://arxiv.org/abs/2607.07430,https://arxiv.org/abs/2607.07430,,,"Apple Vision Pro teleoperation framework coupling egocentric robot visual feedback, voice-controlled locomotion, wrist/finger retargeting, and recording of RGB, commands, joint states, hand motion, and eye gaze for a Unitree H1",digital-teleoperation; humanoid-robotics; dexterous-manipulation; human-robot-interaction; multimodal-data-collection,egocentric-video; vr; gaze; speech; hand-motion; robot-actions
ACE-Brain-0.5,model,2026-07,arXiv,partial,adjacent,2026,https://arxiv.org/abs/2607.04426,https://huggingface.co/ACE-Brain/ACE-Brain-0.5-8B,https://github.com/DAXIAORobotics/ACE-Brain-0.5,,"Unified 8B-class Qwen3-VL-derived embodied backbone spanning spatial perception, planning, navigation/manipulation action generation, progress monitoring, and self-improvement, evaluated across 15 benchmarks",embodied-ai; vla; spatial-reasoning; navigation; manipulation; progress-estimation,robot-camera-video; images; language-instructions; robot-actions; spatial-memory
HeadRoom,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.08083,https://arxiv.org/abs/2607.08083,,,Edge-deployable smart-glasses pipeline that estimates visual and auditory channel availability from egocentric video and audio; a controlled 25-participant study reports faster responses when notifications are routed to the less-loaded channel,wearable-interaction; wearable-assistants; cognitive-state-aware-assistance; edge-ai; on-device-perception,egocentric-video; audio; smart-glasses
VEGAS,benchmark,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.08489,https://arxiv.org/abs/2607.08489,,,Training-free Video caption Evaluation via GAze Score metric plus an evaluation dataset of egocentric activities and instructional slides paired with synchronized gaze and reference captions; the abstract does not report the dataset size,captioning; gaze; attention-modeling; video-text-retrieval; evaluation,egocentric-video; gaze; captions; annotations
CI-HOI / DEHOI,benchmark,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.08514,https://arxiv.org/abs/2607.08514,,,Cue-Isolated HOI evaluation and DEHOI testbed separating hand- and object-related observations through inpainting to diagnose whether egocentric video-language models use manipulation and object-transformation cues rather than contextual shortcuts,hand-object-interaction; action-recognition; object-state-change; evaluation; robot-manipulation,egocentric-video; hand; objects; action-labels; synthetic-video
Whareformer,model,2026-07,ECCV 2026,open,egocentric,2026,https://arxiv.org/abs/2607.08537,https://jacobchalk.github.io/Whareformer/,https://github.com/JacobChalk/Whareformer,MIT,"First learning-based solution for online Out of Sight, Not out of Mind tracking, trained on 56 videos and evaluated on 260 long videos from EPIC-KITCHENS-100, IT3DEgo, and HD-EPIC with updatable appearance and 3D-location memory",long-term-object-tracking; object-reidentification; memory; 3d-understanding; redetection,egocentric-video; object-tracks; 3d-scene; spatial-memory
EgoWAM,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.08436,https://gatech-rl2.github.io/egowam.github.io/,,,"Controlled human-robot co-training study comparing pixel, DINO, and 3D-flow world targets on three real-world bimanual tasks; DINO reports up to 4x gains in out-of-distribution object/scene generalization and 3D flow reports 20-30% in-domain gains",world-action-models; robot-learning; bimanual-manipulation; cross-embodiment-transfer; compositional-generalization,egocentric-video; robot-camera-video; robot-actions; 3d-flow; dense-features
BiNoMaP,model,2025-09,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2509.21256,https://hnuzhy.github.io/projects/BiNoMaP/,,,"Category-level bimanual non-prehensile manipulation primitives learned from hand trajectories extracted from egocentric video demonstrations, geometrically refined and transferred across two dual-arm robot platforms",bimanual-manipulation; retargeting; robot-learning; human-demonstrations; cross-embodiment-transfer,egocentric-video; hand-motion; robot-actions; articulated-objects
GIRAF,model,2026-07,CVPR 2026 HuMoGen Workshop,watch,adjacent,2026,https://arxiv.org/abs/2607.07880,https://arxiv.org/abs/2607.07880,,,"Text-conditioned diffusion model for coordinated full-body interaction with articulated objects, using object-centric contact representations, mixed locomotion/interaction training, and contact-based augmentation",human-motion; motion-generation; articulated-objects; hand-object-interaction; simulation,human-motion; articulated-objects; hand-object-contact; motion-descriptions
Video-Action Generalization Gap / Temporal Ratio,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.08127,https://umishra.me/temporal-ratio/,,,"Systematic video-action model study introducing Temporal Ratio to measure reliance on future latent rollouts, plus inference-time adaptive guidance that narrows compositional OOD gaps on LIBERO and real-world robot tasks",world-action-models; vla; compositional-generalization; action-evaluation; robot-learning,generated-video; latent-video; robot-actions; robot-camera-video
LEEVLA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.08182,https://arxiv.org/abs/2607.08182,,,VLA architecture combining drift-guided dynamic prioritization of task-relevant visual regions with structured latent feature-flow generation for manipulation benchmarks,vla; robot-learning; manipulation; spatial-reasoning,robot-camera-video; robot-actions; latent-actions
LingBot-VA 2.0,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.08639,https://arxiv.org/abs/2607.08639,,,"Native video-action foundation model with a semantic visual-action tokenizer, causal pretraining, sparse mixture-of-experts backbone, and asynchronous closed-loop inference for few-shot robot manipulation",world-action-models; vla; robot-learning; manipulation; efficient-inference,robot-camera-video; robot-actions; latent-video; generated-video
TFP,model,2026-07,RSS 2026 SemRob Workshop,watch,adjacent,2026,https://arxiv.org/abs/2607.08283,https://arxiv.org/abs/2607.08283,,,"Temporally Conditioned Memory-Fusion Policy with a 3.3B VLA backbone and event-sensitive episode belief, reporting 98.75% on LIBERO, 93.77% on LIBERO-plus, and up to 75% on MIKASA ShellGameTouch",vla; long-horizon-memory; robot-learning; manipulation,robot-camera-video; robot-actions; memory
AnyDexRT,toolkit,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.08341,https://chenxi-wang.github.io/projects/anydexrt/,,,"Calibration-free dexterous-hand retargeting that combines self-supervised fingertip correspondence, few-shot human guidance, and contact-aware pinch refinement across multiple hands and real teleoperation tasks",retargeting; digital-teleoperation; dexterous-manipulation; human-demonstrations,hand-pose; robot-actions; hand-object-contact
Harness VLA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.08448,https://arxiv.org/abs/2607.08448,,,"Memory-guided agentic harness that composes a frozen VLA as a retryable contact-rich primitive with fixed analytic primitives, reporting gains on LIBERO-Pro, RoboCasa365, and RoboTwin C2R",vla; long-horizon-memory; manipulation; action-planning; robot-learning,robot-camera-video; robot-actions; memory
FabriVLA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.08575,https://arxiv.org/abs/2607.08575,,,"Compact 1B-parameter VLA combining InternVL3.5 with a flow-matching action head, reporting a 90.0% tier-average success rate across Meta-World MT50",vla; manipulation; robot-learning; efficient-inference,robot-camera-video; robot-actions; language-instructions
DexVerse,benchmark,2026-07,arXiv,partial,adjacent,2026,https://arxiv.org/abs/2607.08751,https://ycyao216.github.io/DexVerse.site/,https://github.com/ycyao216/DexVerse,BSD-3-Clause,"Modular dexterous-manipulation benchmark with 100 tasks, three robot arms, six dexterous hands, configurable visual variation, a VR teleoperation interface, and 3,180 reported demonstrations",benchmark; dexterous-manipulation; bimanual-manipulation; vla; cross-embodiment-transfer,robot-camera-video; depth; point-clouds; robot-actions; demonstrations; simulation
LingBot-Video,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.07675,https://arxiv.org/abs/2607.07675,,,"Mixture-of-Experts DiT video pretraining paradigm for embodied intelligence, using a data-profiling engine that augments internet videos with robot-oriented manipulation, navigation, and egocentric-perspective footage plus physically grounded reward alignment",foundation-video; world-modeling; embodied-ai; robot-learning; video-generation; efficient-inference,video; generated-video; egocentric-video; robot-camera-video; robot-actions
Ego3DLM / Ego-Human Motion Prediction,model,2026-07,ECCV 2026,watch,egocentric,2026,https://arxiv.org/abs/2607.07001,https://arxiv.org/abs/2607.07001,,,"3D-aware LLM for egocentric human motion prediction that jointly decodes past pose, future pose, past narration, and future narration from three-point tracking, 3D scene features, and egocentric video; evaluated on Nymeria",human-motion; pose; trajectory-prediction; motion-language; future-prediction; ar-vr,egocentric-video; 3d-scene; 3d-human-motion; language; trajectories
CoMind,dataset,2026-07,ECCV 2026,watch,egocentric,2026,https://arxiv.org/abs/2607.06691,https://comind.ethz.ch/,,,"Dual-view egocentric and exocentric collaborative-cooking dataset with two synchronized head-mounted cameras, two exocentric views, high-quality audio, gaze tracking, 3D scene scans, object scans, action/hand/object/gaze annotations, and benchmarks for joint attention, socially conditioned object-interaction anticipation, and collaborative handover",social-reasoning; theory-of-mind; visual-grounding; action-anticipation; interaction-understanding,egocentric-video; exocentric-video; audio; gaze; scene-scans; object-meshes; 6dof; annotations
LingBot-World 2.0,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.07534,https://arxiv.org/abs/2607.07534,,,"Interactive world simulator update with unbounded interaction horizon, a real-time distilled variant for 720p 60 FPS streams, richer action/event controls, agentic pilot/director harnesses, and paired 14B/1.3B models",world-modeling; simulation; video-generation; embodied-ai,generated-video; language-instructions; simulation
RoboDojo,benchmark,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.04434,https://arxiv.org/abs/2607.04434,,,"Unified sim-and-real benchmark for generalist robot manipulation with 42 simulation tasks, 18 real-world tasks, five capability dimensions, RoboDojo-RealEval remote evaluation, XPolicyLab integration, 30 evaluated policies, and a public leaderboard",benchmark; evaluation; robot-manipulation; vla; domain-generalization,robot-camera-video; robot-actions; simulation; evaluation
LaMem-VLA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.07608,https://arxiv.org/abs/2607.07608,,,"Latent-memory-native VLA framework with curator, seeker, condenser, and weaver modules that reconstruct short- and long-term historical experience into latent memory tokens for long-horizon robotic manipulation",vla; robot-learning; manipulation; long-horizon-memory; memory-reasoning,robot-camera-video; robot-actions; latent-actions; memory
TouchWorld,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.07287,https://arxiv.org/abs/2607.07287,,,"Predictive-and-reactive tactile foundation model for dexterous manipulation that separates vision-language planning, tactile world-model prediction, visuo-tactile goal-conditioned action generation, and high-frequency tactile residual refinement",tactile-learning; dexterous-manipulation; world-modeling; robot-learning; contact-rich-manipulation,robot-camera-video; tactile; proprioception; robot-actions
WAM-TTT,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.06988,https://arxiv.org/abs/2607.06988,,,Test-time training framework that steers frozen world-action models from unlabeled raw human videos by adapting a lightweight memory through self-supervised video prediction and paired human-robot meta-training,world-action-models; test-time-scaling; robot-learning; human-demonstrations; manipulation,demonstration-video; robot-camera-video; robot-actions; memory
EAGOR,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.06165,https://arxiv.org/abs/2607.06165,,,"Training-free geometry-aware framework for 360-degree directional reasoning that maintains spherical-harmonic belief fields over target directions and reports gains on HOS, OSR-Bench, and legged-robot navigation",spatial-reasoning; navigation; embodied-reasoning; visual-grounding,360-video; robot-observations; maps
VLA Models Review,survey,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.06706,https://arxiv.org/abs/2607.06706,,,"Survey of 183 VLA contributions from 2017-2026 across architectures, training recipes, action representations, bimanual coordination, UAV navigation/control, language grounding, memory, and world models",survey; vla; robot-learning; bimanual-manipulation; navigation,survey; robot-camera-video; robot-actions
NativeMEM,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.06678,https://arxiv.org/abs/2607.06678,,,"Memory-compression VLA policy that repurposes the VLA vision encoder to compress each historical frame from each camera view into a single memory token, improving long-horizon manipulation with low latency",vla; robot-learning; manipulation; long-horizon-memory; efficient-inference,robot-camera-video; robot-actions; memory
Lift3D-VLA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.06564,https://arxiv.org/abs/2607.06564,,,"3D geometry- and dynamics-aware VLA that lifts pretrained 2D models to point-cloud reasoning, uses geometry-centric masked autoencoding, and reports results across 22 simulated and 8 real-world manipulation tasks",vla; robot-learning; 3d-understanding; manipulation; spatial-reasoning,robot-camera-video; point-clouds; robot-actions
Pelican-VLA 0.5,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.06655,https://arxiv.org/abs/2607.06655,,,"Unified VLA report integrating vision-language understanding, future-frame generation, and action prediction, with learnable reasoning slots that induce manipulation-centric attention across unseen scenes and embodiments",vla; robot-learning; manipulation; future-prediction; video-generation,robot-camera-video; generated-video; robot-actions; language-instructions
ActionCache,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.06370,https://arxiv.org/abs/2607.06370,,,"Training-free cache and refinement method for accelerating flow-matching VLA action heads, reusing past intermediate actions through compact multimodal keys and reporting up to 11.75x and 34.43x inference acceleration on representative VLA models",vla; robot-learning; efficient-inference; action-evaluation; manipulation,robot-camera-video; robot-actions; latent-actions
Smooth Operator / SBR,toolkit,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.07491,https://arxiv.org/abs/2607.07491,,,"Sampling-Based Retargeter for low-jitter real-time kinematic hand retargeting, evaluated in simulation and an 18-participant real-world user study over three complex manipulation tasks",retargeting; digital-teleoperation; human-demonstrations; dexterous-manipulation; robot-learning,hand-pose; trajectories; robot-actions
EgoPolice,benchmark,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.06468,https://arxiv.org/abs/2607.06468,,,"Benchmark of real egocentric police-civilian interactions from public body-worn camera footage, with second-by-second high-stakes action labels plus classification and multiple-choice QA tasks",benchmark; action-recognition; video-qa; safety; high-stakes-egocentric-understanding,egocentric-video; body-worn-camera; annotations; qa; action-labels
RynnWorld-4D,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.06559,https://arxiv.org/abs/2607.06559,,,"4D embodied world model for robotic manipulation that co-generates future RGB, depth, and optical flow; reports Rynn4DDataset 1.0 with 254.4M+ frames across egocentric human and robotic manipulation videos",world-modeling; robot-learning; manipulation; vla; action-conditioned-generation,egocentric-video; robot-camera-video; rgb-d; optical-flow; generated-video; robot-actions
RynnWorld-Teleop,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.06558,https://arxiv.org/abs/2607.06558,,,"Digital-teleoperation system where an operator hand-pose stream drives a robot-centric generative world model to synthesize high-fidelity egocentric videos and action labels, with reported 40+ FPS single-pass generation",digital-teleoperation; robot-learning; imitation-learning; world-modeling; data-generation,hand-pose; generated-video; egocentric-video; robot-actions
LingBot-VLA 2.0,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.06403,https://arxiv.org/abs/2607.06403,,,"VLA system update with a redesigned data pipeline and about 60K hours of pretraining data, including 50K hours of robot trajectories across 20 robot configurations and 10K hours of egocentric human videos",vla; robot-learning; manipulation; whole-body-control; world-modeling,egocentric-video; robot-camera-video; robot-actions; language-instructions; depth
CGGS,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.03819,https://arxiv.org/abs/2607.03819,,,"Text-to-3D framework for ego-centric 3D scene generation using a consistency-augmented ego-centric generator, depth/layout decoration from optical flow and point tracks, and geometric Gaussian refinement",text-to-3d; 3d-scene-generation; world-modeling; view-consistency; spatial-reasoning,text; generated-3d; generated-video; depth; optical-flow
EgoInertia-MI,benchmark,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.03934,https://arxiv.org/abs/2607.03934,,,Multimodal benchmark dataset with synchronized egocentric video and wearable IMU for motor-impairment analysis over 19 upper- and lower-body activities and simulated no/mild/severe impairment levels,motor-impairment-assessment; activity-recognition; sensor-fusion; health-ai; wearable-sensing,egocentric-video; imu; wearable-sensing; action-labels
SAGE Action-Gaze,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.04017,https://arxiv.org/abs/2607.04017,,,Synchronized Action-GazE framework for joint recognition and anticipation of human-object interaction and gaze patterns across egocentric close-up and exocentric contextual settings,gaze-prediction; action-anticipation; hand-object-interaction; human-behavior-understanding,egocentric-video; exocentric-video; gaze; action-labels
R3D-Bench / R3D,benchmark,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.02921,https://arxiv.org/abs/2607.02921,,,"Quantitative 3D spatial-reasoning benchmark for egocentric wearables with 3,033 questions over 57 Aria Digital Twin RGB-D video sequences, plus a model-agnostic 3D spatial tool-calling baseline",spatial-reasoning; 3d-understanding; video-qa; ar-wearables; benchmark,egocentric-video; rgb-d; 3d; qa; object-segmentation
S-EMBER,benchmark,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.02689,https://arxiv.org/abs/2607.02689,,,"Streaming Egocentric Memory Benchmark with 3,141 Ray-Ban Meta smart-glasses videos totaling 388 hours and 9,448 QA pairs requiring grounded temporal evidence under causal streaming retrieval",episodic-memory; streaming-video-understanding; retrieval; video-qa; wearable-assistants,egocentric-video; smart-glasses; qa; temporal-localization; streaming-video
Embodied.cpp,toolkit,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.02501,https://arxiv.org/abs/2607.02501,,,"Portable C++ inference runtime for embodied VLA and WAM deployment, with layered adapters, sequence builders, backbone execution, head plugins, and robot-interface deployment support",vla; world-action-models; inference-runtime; robot-deployment; edge-ai,robot-camera-video; robot-actions; embodied-models
EVA-Client,toolkit,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.02646,https://arxiv.org/abs/2607.02646,,,"Unified real-robot client for embodied policy deployment, data collection, and evaluation, with decoupled robot backends, inference strategies, transport middleware, and training-ready rollout logging",robot-deployment; data-collection; evaluation; vla; manipulation,robot-camera-video; robot-actions; logs
InternVLA-A1.5,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.04988,https://arxiv.org/abs/2607.04988,,,"Robot manipulation VLA combining a native VLM backbone, continuous action expert, subtask/VQA supervision, and latent foresight tokens distilled from a frozen video-generation model",vla; robot-learning; manipulation; future-prediction; compositional-generalization,robot-camera-video; robot-actions; language-instructions; latent-video
SIEVE,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.06442,https://arxiv.org/abs/2607.06442,,,"Structure-aware data-selection method for VLA imitation learning that discovers visuo-motor primitives, allocates selection budgets to composition patterns, and selects medoid trajectories",vla; imitation-learning; data-selection; robot-learning,robot-camera-video; robot-actions; trajectories
SVA / Look Before You Leap,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.03751,https://arxiv.org/abs/2607.03751,,,"Search, Value, and Act framework that distills Monte-Carlo tree-search rollouts into a lightweight action evaluator for frozen VLA policies, improving test-time action selection without changing the backbone",vla; robot-learning; action-evaluation; test-time-scaling; planning,robot-camera-video; robot-actions; simulation
HoyerChou Egocentric Bimanual Videos,dataset,2026-01,Hugging Face,open,egocentric,2026,https://arxiv.org/abs/2501.14208,https://huggingface.co/datasets/HoyerChou/EgocentricVideos,,apache-2.0,"YOTO/YOTO++ bimanual manipulation release with three raw first-person task videos, 119 segmented short demonstrations, 241 rows, about 973 MB total, and extracted MANO hand meshes, 3D keypoints, camera calibration, and camera/world transforms",bimanual-manipulation; robot-learning; imitation-learning; hand-object-interaction; hand-tracking,egocentric-video; hand-meshes; hand-pose; camera-pose; annotations
Future-Privileged Causal Ego Gaze,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.01437,https://arxiv.org/abs/2607.01437,,,Controlled causal egocentric gaze-estimation study using a future-aware training branch that is discarded at inference; evaluates EGTEA Gaze+ and Ego4D and finds peak gains around 1.7-3.3 seconds of future context,gaze-prediction; streaming-video-understanding; ar-wearables; attention-modeling,egocentric-video; gaze; streaming-video
NormAct,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.27826,https://arxiv.org/abs/2606.27826,,,"Embodied planning benchmark for hidden social-norm compliance in egocentric environments, reporting Goal Achievement, Norm Compliance, and Task Success and showing a large norm-grounding gap in frontier MLLMs",normative-action-reasoning; action-planning; embodied-reasoning; safety; benchmark,egocentric-observations; language-instructions; annotations
Astra,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.06476,https://arxiv.org/abs/2606.06476,,,Agentic spatial-reasoning framework coupling an RL-trained VLM policy with a view-consistent world simulator to request imagined novel-view evidence from limited egocentric observations,spatial-reasoning; agentic-reasoning; world-modeling; view-synthesis; multimodal-reasoning,egocentric-observations; generated-video; text; camera-motion
Motion Focus Recognition,model,2026-01,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2601.07154,https://arxiv.org/abs/2601.07154,,,"Real-time motion-focus recognition for fast-moving egocentric video, estimating wearer locomotion intention from arbitrary first-person video with camera-pose foundation-model features and sliding-batch inference",activity-recognition; motion-estimation; edge-ai; efficient-inference,egocentric-video; camera-motion; motion
CIVIL Lifelog Retrieval,model,2025-10,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2510.04010,https://arxiv.org/abs/2510.04010,,,Captioning-Integrated Visual Lifelog retrieval system that converts wearable-camera lifelog images into first-person experience captions before text-query retrieval,lifelogging; personalized-retrieval; memory-assistance; captioning,lifelogging-images; captions; text
THOR,model,2025-07,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2507.06442,https://arxiv.org/abs/2507.06442,,,"Thermal-guided adaptive RGB sampling for wearable hand-object activity monitoring, validated with 14 in-the-wild participants and Ego4D, using about 3% of RGB data while preserving activity segments",hand-object-interaction; activity-recognition; wearable-sensing; efficient-inference; healthcare-assistance,egocentric-video; thermal; object-crops
Physical Activity VLM Annotation,model,2025-05,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2505.03374,https://arxiv.org/abs/2505.03374,,,"Compares VLMs and discriminative models for labeling sedentary, light, and moderate-to-vigorous activity from Autographer wearable-camera images across Oxfordshire and Sichuan free-living studies",activity-recognition; lifelogging; wearable-sensing; domain-generalization,egocentric-images; action-labels
Head+Cane Camera Navigation,model,2025-04,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2504.19345,https://arxiv.org/abs/2504.19345,,,"Systematic comparison of synchronized head- and cane-mounted cameras for blind last-mile navigation, evaluating SLAM localization and NeRF-style 3D reconstruction across five real environments",blind-assistance; navigation; scene-reconstruction; sensor-fusion,egocentric-video; wearable-sensing; scene-reconstruction
Audio-Narrated FPV Domain Generalization,model,2024-09,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2409.09611,https://arxiv.org/abs/2409.09611,,,"Multimodal first-person action-recognition framework using motion, audio, appearance, audio narrations, and audio-visual consistency ratings to improve domain generalization on ARGO1M",action-recognition; audio-visual; domain-generalization; multimodal-generalization,first-person-video; audio; narrations; motion
Smart-Glasses Engagement Prediction,dataset,2024-09,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2409.09135,https://arxiv.org/abs/2409.09135,,,Smart-glasses dyadic-conversation dataset from 34 participants with self-reported engagement ratings and a multimodal-transcript LLM fusion strategy for behavioral reasoning,social; engagement-detection; multimodal-conversation; behavior-understanding,smart-glasses; audio; transcripts; annotations
Egocentric 360 VideoQA for Visual Impairment,dataset,2024-05,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2405.19794,https://arxiv.org/abs/2405.19794,,,Visual-question-answering dataset captured with a 360-degree egocentric wearable camera for multiple real-life obstacles faced by people with visual impairments,assistive-videoqa; 360-video; blind-assistance; video-qa,first-person-360-video; qa; annotations
AIM-2 Food Ingestion Environment,model,2024-05,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2405.07827,https://arxiv.org/abs/2405.07827,,,Two-stage fine-tuning and transfer-learning method for recognizing ingestion environment from AIM-2 egocentric wearable-camera images in the UA Free Living Study,nutrition-estimation; activity-recognition; lifelogging; wearable-sensing,egocentric-images; nutrition-labels
LIME,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.02417,https://arxiv.org/abs/2607.02417,,,"Language-conditioned camera-motion generator mined from egocentric video, pairing free-form intents and observation-gain descriptions with relative SE(3) target poses for next-view prediction",active-perception; spatial-reasoning; robot-learning; vla; navigation,egocentric-video; camera-motion; language-instructions; robot-camera-video
LongEgoRefer,benchmark,2026-07,ECCV 2026,open,egocentric,2026,https://arxiv.org/abs/2607.02096,https://github.com/shunya-kato/LongEgoRefer,https://github.com/shunya-kato/LongEgoRefer,,"Long-form Ego4D referring-expression benchmark with 1,498 expressions over videos averaging 45 minutes, requiring temporal and spatial localization of sparse referred objects",referring-expression-comprehension; temporal-grounding; object-grounding; visual-grounding; benchmark,egocentric-video; text; object-boxes; annotations
HandsOnWorld,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.02075,https://arxiv.org/abs/2607.02075,,,Unconstrained egocentric video generation with camera-disentangled hand control; introduces EgoVid-Pro with 103K clips and roughly 12M frames of protagonist-only 3D hand trajectories,video-generation; hand-object-interaction; world-modeling; 3d-hand-pose,egocentric-video; hand-trajectories; camera-motion; generated-video
SG-Ego / GLEN,dataset,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.02425,https://francescapistilli.github.io/GLEN,,,"Large-scale Ego4D spatio-temporal scene-graph annotation set plus GLEN graph model for scene-evolution reasoning, retrieval, and activity-driven graph-edit forecasting",structured-reasoning; action-understanding; scene-graph-qa; temporal-reasoning,egocentric-video; scene-graphs; action-labels; text
Wearable Gait MoCap with Shank-Mounted Ego Cameras,dataset,2026-07,Scientific Data 2026,watch,egocentric,2026,https://doi.org/10.1038/s41597-026-07657-7,https://doi.org/10.1038/s41597-026-07657-7,,,Wearable motion-capture dataset for gait analysis using IMUs and shank-mounted egocentric cameras,activity-recognition; human-motion; sensor-fusion; wearable-sensing,egocentric-video; imu; motion; wearable-sensing
TAP VLA,model,2026-07,ICML 2026,watch,adjacent,2026,https://arxiv.org/abs/2607.02466,https://arxiv.org/abs/2607.02466,,,Task-Agnostic Pretraining framework that learns transferable motor priors from unlabeled interaction data before grounding them with limited language-instruction data,vla; robot-learning; representation-learning; imitation-learning,robot-camera-video; robot-actions; language-instructions
The Moving Eye,model,2026-07,IROS 2026,watch,adjacent,2026,https://arxiv.org/abs/2607.02322,https://arxiv.org/abs/2607.02322,,,"Hybrid dynamic data-collection strategy for VLA spatial generalization, using a dual-arm setup where one arm manipulates while the other acts as a moving environmental camera",vla; robot-learning; active-perception; data-collection; spatial-reasoning,robot-camera-video; robot-actions; multiview-video
Bridge-WA,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.02195,https://hcplab-sysu.github.io/BRIDGE-WA,,,"Lightweight world-action framework that distills future-change tokens, change maps, and motion-flow maps into a VLA action transformer",vla; robot-learning; world-modeling; future-state-prediction,robot-camera-video; robot-actions; generated-video
VLA-Corrector,model,2026-07,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.01804,https://arxiv.org/abs/2607.01804,,,"Lightweight detect-and-correct inference layer for action-chunked VLA policies, using visual feature evolution to detect open-loop deviations and adapt action horizons",vla; robot-learning; efficient-inference; manipulation-safety,robot-camera-video; robot-actions
VLA-Arena,benchmark,2025-12,ICML 2026,watch,adjacent,2025,https://arxiv.org/abs/2512.22539,https://arxiv.org/abs/2512.22539,,,"Open-source VLA benchmark with 170 structured tasks and decoupled task, language, and visual perturbation axes for measuring model capability frontiers",vla; benchmark; evaluation; manipulation-safety,robot-camera-video; language-instructions; robot-actions
VLSA / AEGIS,model,2025-12,IROS 2026,watch,adjacent,2025,https://arxiv.org/abs/2512.11891,https://arxiv.org/abs/2512.11891,,,Vision-Language-Safe Action architecture with a plug-and-play control-barrier-function safety layer and SafeLIBERO benchmark for safety-critical manipulation,vla; manipulation-safety; safety; robot-learning; benchmark,robot-camera-video; language-instructions; robot-actions
H-Tac / TTP,dataset,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.01067,https://arxiv.org/abs/2607.01067,,,"Human-Centric Transferable Tactile Pre-Training introduces H-Tac, a 160-hour tactile-action dataset with egocentric human videos over 300+ tasks and 135K episodes, plus tactile pretraining for fine-grained robotic manipulation",tactile-learning; robot-learning; dexterous-manipulation; human-demonstrations; vla,egocentric-video; tactile; force; human-demonstrations; robot-actions
Imprint,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.00696,https://arxiv.org/abs/2607.00696,,,"Online interaction-centric memory compression for long-horizon egocentric QA, evaluated on seven-day EgoLifeQA with improved QA accuracy, evidence grounding, memory footprint, and retrieval latency",episodic-memory; long-context-qa; egocentric-video-qa; retrieval; memory-reasoning,egocentric-video; qa; text
EgoGapBench,benchmark,2026-07,arXiv,open,egocentric,2026,https://arxiv.org/abs/2607.00547,https://github.com/jhCOR/EgoGapBench,https://github.com/jhCOR/EgoGapBench,,"Diagnostic benchmark for egocentric action selection in multi-agent scenes, isolating whether models choose actions from the camera-wearer's perspective rather than copying visible agents",action-reasoning; action-selection; egocentric-vqa; multi-agent-memory; benchmark,egocentric-video; qa; action-labels; annotations
EgoSafetyBench,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.00218,https://arxiv.org/abs/2607.00218,,,"1,200 egocentric robot-view safety scenarios annotated at half-second granularity, with situational and misleading in-scene text tracks for evaluating streaming VLM safety guards",safety; vlm-evaluation; streaming-video-understanding; scene-text; benchmark,egocentric-video; robot-camera-video; scene-text; annotations
RetailSMV,dataset,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.00310,https://arxiv.org/abs/2607.00310,,,"32,105 captioned synchronized retail clips from five supermarkets with staff-perspective egocentric and exocentric capture, used to compare LoRA adaptation of foundation video world models",world-modeling; video-generation; domain-adaptation; retail-understanding; ego-exo,egocentric-video; exocentric-video; captions
Human-as-Humanoid,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.32009,https://zgc-embodyai.github.io/Human-as-Humanoid,,,"Human-to-humanoid supervision framework that uses synchronized ego-exo videos, exocentric motion recovery, staged IK, and FK-aware supervision to convert human demonstrations into 60-DoF humanoid action chunks",humanoid-control; whole-body-manipulation; imitation-learning; cross-embodiment-transfer; vla,egocentric-video; exocentric-video; human-demonstrations; humanoid-motion; robot-actions
SkillSpotter,model,2026-06,ECCV 2026,open,egocentric,2026,https://arxiv.org/abs/2606.31127,https://github.com/eth-siplab/SkillSpotter,https://github.com/eth-siplab/SkillSpotter,,"Pose-aware multi-view skilled-action detection and grading for Ego-Exo4D proficiency demonstrations, using adaptive temporal suppression, gated 3D body-pose fusion, and bidirectional ego-exo attention",skill-assessment; action-detection; proficiency; ego-exo; pose,egocentric-video; exocentric-video; 3d-human-pose; action-labels
ViDiHand,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.30308,https://vidihand.github.io,https://github.com/NTUYWANG103/ViDiHand,,"4D two-hand motion reconstruction from full-frame egocentric video using video-diffusion representations and a hand-overlay rendering objective, evaluated on ARCTIC, HOT3D, and HOI4D",3d-hand-pose; hand-tracking; hand-reconstruction; hand-object-interaction,egocentric-video; hand-pose; hand-meshes
Controllable Egocentric Video Generation,model,2026-03,ECCV 2026,watch,egocentric,2026,https://arxiv.org/abs/2603.11755,https://arxiv.org/abs/2603.11755,,,"Occlusion-aware sparse 3D hand-joint control for realistic egocentric hand-object video generation, with an automated pipeline yielding 1M egocentric clips paired with precise hand trajectories",video-generation; hand-object-interaction; world-modeling; 3d-hand-pose,egocentric-video; hand-pose; generated-video; trajectories
EgoVITA,model,2025-11,ECCV 2026,watch,egocentric,2025,https://arxiv.org/abs/2511.18242,https://arxiv.org/abs/2511.18242,,,"Plan-then-verify framework for egocentric video reasoning, using GRPO with visual-grounding and cross-perspective consistency rewards over a 52K-sample training setup",egocentric-video-reasoning; planning; temporal-reasoning; grounded-reasoning,egocentric-video; qa; text
OmniView-Space,model,2026-07,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2607.00881,https://arxiv.org/abs/2607.00881,,,"Spatial-reasoning framework that re-anchors reconstructed geometry into query-aligned visual cognitive maps and textual spatial graphs, then trains tool-guided egocentric reasoning with ego-frame rewards",spatial-reasoning; 3d-understanding; grounded-reasoning; mllm,egocentric-images; 3d; scene-graphs; text
LongVQUBench,benchmark,2026-07,ECCV 2026,watch,adjacent,2026,https://arxiv.org/abs/2607.01086,https://arxiv.org/abs/2607.01086,,,"Long-term video-quality understanding benchmark with 1,200+ videos and 1,500 questions spanning movies, documentaries, surveillance, egocentric recordings, and animation",benchmark; vlm-evaluation; long-video-understanding,video; qa; annotations
WorldOdysseyBench,benchmark,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.31672,https://arxiv.org/abs/2606.31672,,,"600+ open-world interactive world-model test cases across nature, urban, and indoor scenes, using first/third-person views and 10-60s WASD interaction to test action, vision, physics, and memory stability",world-modeling; benchmark; long-horizon-reasoning; evaluation,video; generated-video; 3d; trajectories
MM-Nav,model,2025-10,arXiv,watch,adjacent,2025,https://arxiv.org/abs/2510.03142,https://pku-epic.github.io/MM-Nav-Web/,,,"Multi-view VLA navigation model with 360-degree observations, trained from synthetic RL expert data for reaching, squeezing, and avoiding, with real-world navigation experiments",vla; navigation; robot-learning; evaluation,robot-camera-video; synthetic-rendering; language-instructions; trajectories
BinaryTracking / GangnamLoop,benchmark,2026-06,arXiv,open,egocentric,2026,https://arxiv.org/abs/2606.16902,https://github.com/ndb796/BinaryTracking,https://github.com/ndb796/BinaryTracking,,"Spatial QA and navigation benchmark for service robots traversing long egocentric routes, with the GangnamLoop multi-trip outdoor route dataset and BinTrack open implementation",spatial-reasoning; navigation; egocentric-video-qa; episodic-memory; benchmark,egocentric-video; trajectory; text; annotations
WT-UMI,toolkit,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.13232,https://arxiv.org/abs/2606.13232,,,"Wearable whole-body tactile UMI interface for humanoid manipulation, aligning human demonstrations and teleoperation through tactile images, contact forces, and end-effector poses",whole-body-manipulation; tactile-learning; imitation-learning; humanoid-control; vla,egocentric-video; tactile; force; human-demonstrations; humanoid-motion
UMI-Bench 1.0,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.10382,https://arxiv.org/abs/2606.10382,,,"Local-first real-robot benchmark dedicated to UMI-style manipulation policies, aligning UMI data collection, scene reset, deployment, logging, and task-factor analysis",benchmark; robot-learning; manipulation; imitation-learning; evaluation,wrist-camera-video; robot-actions; annotations
VISTA UMI,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.04708,https://arxiv.org/abs/2606.04708,,,"Vision-grounded and physics-validated adaptation pipeline for UMI data, combining UMI-VQA for wrist fisheye visual alignment with physics filtering for VLA training",vla; robot-learning; manipulation; data-collection; physical-action-understanding,wrist-camera-video; robot-actions; language-instructions; annotations
HumanoidUMI,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.27239,https://arxiv.org/abs/2606.27239,,,"Robot-free robot-wide demonstration framework using lightweight VR and UMI-inspired grippers to collect sparse human keypoint trajectories and wrist-view observations, train high-level keypoint prediction, retarget to humanoid references, and execute whole-body control",humanoid-control; whole-body-manipulation; loco-manipulation; imitation-learning; data-collection,human-demonstrations; wrist-camera-video; trajectories; robot-actions; humanoid-motion
BRIDGE State-Gated Experts,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.26603,https://arxiv.org/abs/2606.26603,,,Combines handheld UMI demonstrations with targeted teleoperated segments through state-gated diffusion-policy experts for contact-rich manipulation,robot-learning; manipulation; imitation-learning; cross-embodiment-transfer; dexterous-manipulation,wrist-camera-video; robot-actions; human-demonstrations; trajectories
ForceBand,dataset,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.26093,https://forceband-emg.github.io,,,"10-hour multimodal wrist-worn sEMG dataset with egocentric video, IMU, fingertip force, EMG2Force labels, and force-augmented demonstrations for robot manipulation",robot-learning; manipulation; tactile-learning; imitation-learning; sensor-fusion,egocentric-video; emg; imu; force; human-demonstrations
Egocentric Scene Graphs / EgoSG,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.25842,https://arxiv.org/abs/2606.25842,,,Long-form egocentric video VQA framework that converts videos into temporally grounded scene graphs for compact MLLM reasoning on HD-EPIC VQA,egocentric-video-qa; long-video-qa; scene-graph-qa; structured-reasoning,egocentric-video; scene-graphs; qa; text
NetraLink Assistive AI,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.25084,https://arxiv.org/abs/2606.25084,,,"Head-mounted GoPro egocentric benchmark for assistive MLLMs covering object recognition, scene text QA, and multilingual visual reading",assistive-systems; assistive-videoqa; video-qa; wearable-sensing; scene-understanding,egocentric-video; first-person-video; text; qa
Low-Latency Doubly-Correct VLMs,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.25160,https://arxiv.org/abs/2606.25160,,,Rationale-informed pruning study for low-latency VLMs that preserves both prediction correctness and evidence grounding on egocentric video benchmarks,efficient-inference; egocentric-video-reasoning; vlm-evaluation; evaluation,egocentric-video; vlm; annotations
XRZero-G0,toolkit,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.13001,https://github.com/X-Square-Robot/XRZero-G0,https://github.com/X-Square-Robot/XRZero-G0,,"VR and dual-gripper robot-free data-collection system with a reported 2,000-hour robot-free dataset and empirical data-mixing ratios for zero-shot cross-embodiment transfer",robot-learning; data-collection; imitation-learning; cross-embodiment-transfer; manipulation,wrist-camera-video; human-demonstrations; robot-actions; trajectories
EgoTSR,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.10517,https://arxiv.org/abs/2604.10517,,,"Curriculum-based task-oriented egocentric spatiotemporal reasoning model with EgoTSR-Data, a reported 46M-sample training set across CoT supervision, weak tags, and long-horizon sequences",egocentric-video-reasoning; spatial-reasoning; multistep-reasoning; action-planning,egocentric-video; text; annotations
FlexLAM,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.19408,https://arxiv.org/abs/2606.19408,,,"Variable-length latent-action model trained by nested dropout, improving scarce-label action alignment and Ego4D transition reconstruction",latent-action-learning; world-modeling; robot-learning; egocentric-representation,egocentric-video; latent-actions; robot-actions
VisualClaw,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.16295,https://arxiv.org/abs/2606.16295,,,"Real-time personalized multimodal agent that filters streaming first-person frames, evolves a skill bank from failures, and introduces VisualClawArena for workspace-grounded video-agent evaluation",assistant; streaming-assistant; tool-using-agents; video-qa; personalized-qa,first-person-video; video-qa; memory; digital-context
LifeDialBench / EgoMem,benchmark,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.11182,https://arxiv.org/abs/2604.11182,,,Continuous lifelog memory benchmark with EgoMem built on real-world egocentric videos plus LifeMem synthetic community scenarios and online temporal-causality evaluation,lifelogging; personal-memory; long-video-qa; streaming-video-understanding; benchmark,egocentric-video; audio; dialogue; memory
TAIHRI,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.08921,https://arxiv.org/abs/2604.08921,https://github.com/Tencent/TAIHRI,,Task-aware VLM for close-range HRI that localizes task-relevant 3D human keypoints in the egocentric camera coordinate frame,3d-human-pose; hri; spatial-reasoning; human-pose-estimation,egocentric-video; human-pose; 3d-pose; text-instructions
MASS,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.08943,https://arxiv.org/abs/2604.08943,,,Mesh-inellipse aligned deformable surfel splatting for high-fidelity 3D hand reconstruction and rendering from egocentric monocular video,hand-reconstruction; 3d-hand-pose; 3d-reconstruction; hand-object-interaction,egocentric-video; hand-pose; hand-meshes; 3d-gaussian-splatting
E-3DPSM,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.08543,https://arxiv.org/abs/2604.08543,,,"Event-driven continuous pose state machine for event-based egocentric 3D human-pose estimation from head-mounted devices, reporting real-time 80 Hz inference",egocentric-3d-pose; human-pose-estimation; ar-vr; wearable-sensing,event-camera; egocentric-video; 3d-human-pose; body-pose
EgoSMPLX,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.08606,https://github.com/naso06/EgoSMPLX,https://github.com/naso06/EgoSMPLX,,"Prior-guided whole-body human mesh recovery from monocular head-mounted egocentric images, with optimization-based pseudo-GT and released code/annotations",mesh-recovery; body-shape; egocentric-3d-pose; ar-vr,egocentric-images; body-pose; hand-pose; human-avatar
AwareLLM,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.09625,https://arxiv.org/abs/2605.09625,,,"Proactive multimodal assistant ecosystem combining egocentric vision, pupillometry, gaze, posture, heart activity, and LLM reasoning for personalized productivity support",proactive-assistance; internal-state-reasoning; wearable-sensing; personalized-qa,egocentric-video; gaze; eye-tracking; physiological-signals; posture
SpaceMind++,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.09449,https://arxiv.org/abs/2605.09449,,,Video MLLM that builds allocentric cognitive maps from RGB video to reorganize fragmented egocentric observations for spatially consistent 3D reasoning,spatial-reasoning; embodied-reasoning; video-qa; 3d-understanding,egocentric-video; 3d; spatial-memory; vlm
VIGIL,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.08747,https://arxiv.org/abs/2605.08747,,,"Embodied-agent evaluation protocol with egocentric RGB observations and terminal-commitment scoring, separating world-state completion from benchmark success across 1,000 frozen episodes",benchmark; embodied-reasoning; evaluation; task-progress; goal-reasoning,egocentric-video; state-labels; annotations
Egocentric Pedestrian Crossing Intention,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.09142,https://arxiv.org/abs/2606.09142,,,"VLM-based closed-ended VQA formulation for decoding pedestrian crossing intention from short egocentric traffic-safety clips, including zero-shot and parameter-efficient fine-tuning studies",intent-understanding; vqa; safety; trajectory-prediction; wearable-sensing,egocentric-video; qa; annotations
FactCheck LTA,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.14778,https://arxiv.org/abs/2606.14778,,,Feasibility-aware long-term action anticipation using a multi-agent Observe-Plan-Verify loop to reduce physically impossible future action predictions on EPIC-KITCHENS-55 and EGTEA Gaze+,anticipation; long-horizon-memory; action-planning; affordance; procedure-understanding,egocentric-video; action-labels; text
TrAction,model,2026-06,arXiv,open,egocentric,2026,https://arxiv.org/abs/2606.03490,https://github.com/ecker-lab/TrAction,https://github.com/ecker-lab/TrAction,,Sparse 2.5D point-trajectory transformer for action recognition that reduces appearance/background shortcuts and evaluates on EPIC-KITCHENS-100,action-recognition; efficient-inference; motion-estimation; representation-learning,egocentric-video; trajectories
AmbientEye,dataset,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.03774,https://arxiv.org/abs/2606.03774,,,"Smart-glasses pupil-segmentation dataset for passive infrared eye tracking under natural ambient illumination, targeting all-day outdoor use without active IR lighting",gaze; wearable-sensing; segmentation; energy-efficient-smart-glasses,eye-tracking; smart-glasses; annotations
Personal Visual Context Learning / Personal-VCL-Bench,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.10936,https://vision.cs.utexas.edu/projects/PersonalVCL/,,,"Personal Visual Context Learning benchmark for smart-glasses LMMs, testing prompt-time use of wearer-specific visual context over continuous first-person streams",personalized-qa; personal-memory; wearable-sensing; long-context-video; benchmark,egocentric-video; smart-glasses; egocentric-context-labels; qa
GazeMind / CogLoad-Bench,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.05790,https://arxiv.org/abs/2605.05790,,,"Gaze-guided LLM agent and CogLoad-Bench for personalized cognitive-load assessment on smart glasses, reporting 152 participants, 40+ hours, and 10K+ annotations",gaze-reasoning; internal-state-reasoning; assistive-systems; wearable-sensing; benchmark,gaze; eye-tracking; smart-glasses; annotations
SpatialBench,benchmark,2026-05,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2605.27367,https://ropedia.github.io/SpatialBench/,,,"Broad spatial foundation-model benchmark across viewpoints, scene domains, input densities, and hardware constraints, including embodied and egocentric-style spatial reasoning settings",spatial-reasoning; benchmark; embodied-ai; evaluation; 3d-understanding,video; 3d; point-clouds; robot-camera-video; synthetic-rendering
Dietary Behavior Change Receptivity,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.27950,https://arxiv.org/abs/2605.27950,,,Pilot study using free-living AIM-2 wearable-camera egocentric eating images to infer behavior-change receptivity dimensions for just-in-time dietary interventions,lifelogging; nutrition-estimation; healthcare-assistance; behavior-understanding,egocentric-images; lifelogging-images; annotations
E3C Video Generation,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.26316,https://e3c-videogen.github.io/,,,Controllable egocentric video diffusion with 3D environmental memory and ego-exo human-pose control for physically grounded first-person interaction generation,video-generation; world-modeling; 3d-4d-understanding; cross-view; hand-object,egocentric-video; exocentric-video; human-pose; point-clouds; generated-video
Mobile UMI,toolkit,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.20894,https://arxiv.org/abs/2605.20894,,,"Robot-free mobile manipulation demonstration framework with chest-centric global context, wrist-centric local interaction, decoupled kinematics, and latency-aware cross-view diffusion policy",mobile-manipulation; imitation-learning; data-collection; robot-learning; vla,egocentric-video; wrist-camera-video; human-demonstrations; robot-actions
BifrostUMI,toolkit,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.03452,https://arxiv.org/abs/2605.03452,,,"Portable robot-free humanoid demonstration framework using lightweight VR devices, sparse keypoint trajectories, wrist-mounted visual data, and keypoint retargeting",humanoid-control; whole-body-manipulation; imitation-learning; robot-learning; data-collection,wrist-camera-video; hand-trajectories; human-demonstrations; humanoid-motion
UMI-3D,toolkit,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.14089,https://umi-3d.github.io/,,,"Multimodal UMI extension adding low-cost wrist LiDAR, LiDAR-centric SLAM, synchronized sensing, and spatiotemporal calibration to reduce monocular SLAM failures",3d-reconstruction; robot-learning; data-collection; manipulation; egocentric-calibration,wrist-camera-video; point-clouds; calibration; robot-actions
OmniUMI,toolkit,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.10647,https://arxiv.org/abs/2604.10647,,,"Human-aligned multimodal UMI system synchronizing RGB, depth, trajectory, tactile sensing, internal grasp force, and external interaction wrench for contact-rich robot learning",tactile-learning; contact-understanding; imitation-learning; robot-learning; manipulation,wrist-camera-video; depth; tactile; force; robot-actions
UMI-Underwater,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.27012,https://umi-under-water.github.io/,,,Transfers on-land handheld human demonstrations to underwater grasping through depth-based affordance representations plus self-supervised underwater demonstration collection,robot-learning; manipulation; affordance; domain-transfer; imitation-learning,egocentric-video; depth; robot-actions; human-demonstrations
HoMMI,toolkit,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.03243,https://hommi-robot.github.io/,,,"Whole-body mobile manipulation interface learning from robot-free human demonstrations by augmenting UMI with egocentric sensing, relaxed head actions, and cross-embodiment hand-eye policy design",mobile-manipulation; whole-body-manipulation; imitation-learning; active-perception; vla,egocentric-video; wrist-camera-video; human-demonstrations; robot-actions
EgoAdapt Speaker Detection,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.18082,https://arxiv.org/abs/2603.18082,,,"Adaptive framework for egocentric Talking-to-Me speaker detection under missing modalities, head-orientation ambiguity, and background noise in social interaction video",social; audio-visual; first-person-interaction-recognition; robustness,egocentric-video; audio; head-pose; annotations
FlowNar,model,2026-05,ICML 2026,open,egocentric,2026,https://arxiv.org/abs/2606.00620,https://github.com/zeyun-zhong/FlowNar,https://github.com/zeyun-zhong/FlowNar,,"Scalable streaming video narration framework with bounded visual memory via historical-context pruning and Cross Linear Attentive Memory, evaluated on long-form Ego4D, Ego-Exo4D, and EPIC-KITCHENS-100-style narration",narration-generation; streaming-video-understanding; long-video-understanding; video-language,egocentric-video; captions; long-video; streaming-video
egoPPG,model,2025-02,ICCV 2025,watch,egocentric,2025,https://arxiv.org/abs/2502.20879,https://arxiv.org/abs/2502.20879,,,"Heart-rate estimation task for egocentric systems using eye-tracking cameras, adding physiological-state awareness to downstream wearer-behavior understanding",physiological-signals; wearable-sensing; gaze; affect; egocentric-representation,eye-tracking; physiological-signals; smart-glasses
Photoreal Scene Reconstruction from an Egocentric Device,model,2025-06,SIGGRAPH 2025,watch,egocentric,2025,https://arxiv.org/abs/2506.04444,https://www.projectaria.com/photoreal-reconstruction/,,,Project Aria egocentric-device scene reconstruction study showing frame-rate VIO pose is insufficient for pixel-accurate photoreal HDR reconstruction and introducing improved pose/exposure treatment,scene-reconstruction; novel-view-synthesis; ar-vr; 3d-reconstruction,aria; egocentric-video; camera-pose; illumination
Egocentric Event-Based Ping Pong,dataset,2025-06,CVPRW 2025,watch,egocentric,2025,https://arxiv.org/abs/2506.07860,https://arxiv.org/abs/2506.07860,,,Egocentric event-camera table-tennis trajectory-prediction system and dataset for high-speed ball tracking under low latency and motion blur,trajectory-prediction; active-perception; wearable-sensing; benchmark,event-camera; event-stream; egocentric-video; trajectories
Aria Gen 2 Pilot Dataset,dataset,2025-10,arXiv,open,egocentric,2025,https://arxiv.org/abs/2510.16134,https://www.projectaria.com/datasets/,,Project Aria dataset terms,"Incrementally released egocentric multimodal open dataset captured with Aria Gen 2 glasses, covering daily activities such as cleaning, cooking, eating, playing, and outdoor walking",ar-perception; wearable-sensing; everyday-activity; scene-understanding; data-collection,aria; egocentric-video; gaze; imu; audio; calibration
GateFusion,model,2025-12,WACV 2026,watch,egocentric,2025,https://arxiv.org/abs/2512.15707,https://arxiv.org/abs/2512.15707,,,"Hierarchical gated cross-modal fusion model for active speaker detection, improving fine-grained audio-visual interactions on unconstrained and egocentric ASD settings including Ego4D-ASD",audio-visual; social; first-person-interaction-recognition; detection,egocentric-video; audio; annotations
CS3 Collision Sound Source Segmentation,benchmark,2025-11,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2511.13863,https://krantiparida.github.io/projects/cs3.html,,,"Collision Sound Source Segmentation task with EPIC-CS3 and Ego4D-CS3 benchmarks, segmenting objects responsible for collision sounds in egocentric audio-visual clips",audio-visual; sound-understanding; segmentation; hand-object; benchmark,egocentric-video; audio; object-masks; annotations
EAGLE VQL,model,2025-11,AAAI 2026,watch,egocentric,2025,https://arxiv.org/abs/2511.08007,https://arxiv.org/abs/2511.08007,,,"Episodic appearance- and geometry-aware memory for unified 2D/3D visual query localization in egocentric video, targeting Ego4D visual-query localization",visual-query-localization; episodic-memory; object-grounding; 3d-object-detection,egocentric-video; object-crops; 3d; camera-pose
HieraMamba,model,2025-10,CVPR 2026,watch,egocentric,2025,https://arxiv.org/abs/2510.23043,https://vision.cs.utexas.edu/projects/hieramamba/,,,"Hierarchical Anchor-Mamba pooling for long-video temporal grounding, preserving fine temporal structure while improving Ego4D Natural Language Query grounding",temporal-grounding; natural-language-query; video-temporal-grounding; long-video-understanding,egocentric-video; text; long-video
AssemblyHands,dataset,2023-04,CVPR 2023,benchmark,egocentric,2023,https://arxiv.org/abs/2304.12301,https://arxiv.org/abs/2304.12301,,,"3.0M 3D hand-pose annotations sampled from Assembly101, including 490K egocentric images, with an action-classification task using predicted hand poses",3d-hand-pose; hand-object; action-recognition; benchmark,egocentric-images; exocentric-video; hand-pose; annotations
HARMONIC,dataset,2018-07,IJRR 2022,watch,egocentric,2018,https://arxiv.org/abs/1807.11154,https://arxiv.org/abs/1807.11154,,,"Assistive eating HRI dataset from 24 participants with head-mounted egocentric video, eye video/gaze, EMG, joystick commands, robot arm state, third-person stereo video, and derived body/hand/face features",hri; intent-understanding; assistive-systems; sensor-fusion,egocentric-video; gaze; audio; imu; third-person-video; robot-actions
EgoReID Dataset,dataset,2018-12,arXiv,watch,egocentric,2018,https://arxiv.org/abs/1812.09570,https://arxiv.org/abs/1812.09570,,,"First-person person re-identification dataset from three mobile first-person cameras, reporting 900 identities, about 10.2K tracks, 176K detections, and 12-sensor metadata",person-localization; object-reidentification; social; wearable-sensing,first-person-video; annotations; imu
MMG-Ego4D,benchmark,2023-05,CVPR 2023,benchmark,egocentric,2023,https://arxiv.org/abs/2305.07214,https://arxiv.org/abs/2305.07214,https://github.com/facebookresearch/MMG_Ego4D,,"Human-verified Ego4D-derived benchmark for multimodal generalization under missing-modality and cross-modal zero-shot settings over video, audio, and IMU",action-recognition; multimodal-generalization; domain-generalization; benchmark,egocentric-video; audio; imu; annotations
EgoAdapt 2023,benchmark,2023-07,arXiv,benchmark,egocentric,2023,https://arxiv.org/abs/2307.05784,https://arxiv.org/abs/2307.05784,,,"Real-world egocentric action-recognition adaptation benchmark over 50 independent Ego4D user streams and 2,740 action labels, with adaptation gain, online generalization, and hindsight metrics",action-recognition; continual-learning; domain-adaptation; benchmark,egocentric-video; annotations
EPIC-Aff / Multi-label Affordance Mapping,benchmark,2023-09,ICCV 2023,benchmark,egocentric,2023,https://arxiv.org/abs/2309.02120,https://arxiv.org/abs/2309.02120,,,"EPIC-KITCHENS-derived metric, spatial, multi-label affordance annotations for interaction-grounded affordance segmentation and action-centric navigation maps",affordance; segmentation; hand-object; navigation; benchmark,egocentric-video; annotations
VQLoC,model,2023-06,NeurIPS 2023,open,egocentric,2023,https://arxiv.org/abs/2306.09324,https://hwjiang1510.github.io/VQLoC/,,,"Single-stage visual query localization for long-form egocentric video, jointly modeling query-frame and frame-frame correspondences; reports about 20% accuracy improvement and 10x faster inference over prior multi-stage VQL pipelines",visual-query-localization; episodic-memory; object-grounding; tracking,egocentric-video; object-crops; annotations
EgoCOL,model,2023-06,CVPR 2023 Ego4D Challenge,open,egocentric,2023,https://arxiv.org/abs/2306.16606,https://github.com/BCV-Uniandes/EgoCOL,,,"Egocentric camera-pose estimation for Ego4D open-world 3D object localization, using sparse pose reconstruction over video and scan renders",3d-object-detection; visual-query-localization; object-grounding; spatial-localization,egocentric-video; camera-pose; annotations
CocoFormer VQL,model,2022-11,Ego4D Challenge 2022,open,egocentric,2022,https://arxiv.org/abs/2211.10528,https://github.com/facebookresearch/vq2d_cvpr,,,"Conditioned Contextual Transformer for egocentric visual query localization, modeling proposal-set context and improving VQ2D/VQ3D localization through query-conditioned object proposal reasoning",visual-query-localization; object-grounding; detection; tracking,egocentric-video; object-crops; annotations
Negative Frames Matter,model,2022-08,Ego4D Challenge 2022,open,egocentric,2022,https://arxiv.org/abs/2208.01949,https://github.com/facebookresearch/vq2d_cvpr,,,Efficient Ego4D VQ2D solution reducing repeated object-gradient computation and explicitly training with noisy negative/background egocentric frames,visual-query-localization; object-grounding; redetection; benchmark,egocentric-video; object-crops; annotations
SpotEM,model,2023-06,ICML 2023,watch,egocentric,2023,https://arxiv.org/abs/2306.15850,https://vision.cs.utexas.edu/projects/spotem,,,"Efficient episodic-memory search for long egocentric videos, selecting promising query-conditioned video regions and preserving 84-97% of baseline accuracy while computing 10-25% of clip features",episodic-memory; natural-language-query; retrieval; efficient-inference,egocentric-video; natural-language; annotations
CliMer Temporal Grounding,model,2023-10,BMVC 2023,open,egocentric,2023,https://arxiv.org/abs/2310.17395,https://github.com/keflanagan/CliMer,,,Learns temporal sentence grounding from narrated egocentric videos using clip merging and rough narration timestamps instead of dense temporal labels,temporal-grounding; video-language-grounding; narration-generation,egocentric-video; narrations; natural-language
Object-Centric LTA,model,2023-10,WACV 2024,watch,egocentric,2023,https://arxiv.org/abs/2311.00180,https://arxiv.org/abs/2311.00180,,,"Object-prompt representation for long-term action anticipation, extracting task-specific object-centric features from visual-language pretrained models",action-anticipation; object-recognition; video-language-representation,egocentric-video; natural-language; annotations
OWL,model,2022-02,CVPRW 2023,watch,egocentric,2022,https://arxiv.org/abs/2202.04947,https://arxiv.org/abs/2202.04947,,,"Observe, Watch, Listen baseline showing audiovisual temporal context improves egocentric temporal action localization on EPIC-KITCHENS and HOMAGE",temporal-grounding; action-detection; audio-visual,egocentric-video; audio; annotations
Hand-Object Interaction Reasoning,model,2022-01,AVSS 2022,watch,egocentric,2022,https://arxiv.org/abs/2201.04906,https://arxiv.org/abs/2201.04906,,,"Transformer-style interaction unit for reasoning over acting hands, the other hand, and interacted objects in egocentric video",hand-object; action-recognition; relation-reasoning,egocentric-video; hand-trajectories; object-trajectories
Human Hands as Probes,model,2021-12,CVPR 2022,watch,egocentric,2021,https://arxiv.org/abs/2112.09120,https://arxiv.org/abs/2112.09120,,,"Learns interactive object understanding, state-sensitive features, affordance regions, and grasp affordances by observing hands in egocentric video",affordance; hand-object; object-discovery; interaction,egocentric-video; hands; annotations
NeuralDiff,model,2021-10,3DV 2021,watch,egocentric,2021,https://arxiv.org/abs/2110.09936,https://arxiv.org/abs/2110.09936,,,"Triple-stream neural rendering method for separating static background, actor, and moving objects in egocentric video; introduces annotations for dynamic object segmentation on EPIC-KITCHENS",dynamic-scene-understanding; segmentation; 3d-reconstruction; object-motion,egocentric-video; annotations
Head-Motion SSL,model,2021-10,ICCV 2021 EPIC Workshop,watch,egocentric,2021,https://arxiv.org/abs/2110.01680,https://arxiv.org/abs/2110.01680,,,Self-supervised egocentric video representation learning by matching video clips with head-motion signals from IMU sensors embedded in AR/VR devices,representation-learning; activity-recognition; sensor-fusion; ar-vr,egocentric-video; imu
H+O,model,2019-04,CVPR 2019 Oral,watch,egocentric,2019,https://arxiv.org/abs/1904.05349,https://arxiv.org/abs/1904.05349,,,"Unified egocentric RGB model for jointly estimating 3D hand pose, object pose, interaction state, object class, and action class",3d-hand-pose; hand-object; action-recognition; object-recognition,egocentric-video; hand-pose; object-pose
You2Me,model,2019-04,CVPR 2020,watch,egocentric,2019,https://arxiv.org/abs/1904.09882,https://arxiv.org/abs/1904.09882,,,Infers second-person body pose in egocentric video using first- and second-person interaction cues,human-pose; social; interaction; pose,egocentric-video; human-pose
Ego-Surfing,model,2016-06,IEEE TPAMI,watch,egocentric,2016,https://arxiv.org/abs/1606.04637,https://arxiv.org/abs/1606.04637,,,Classic first-person person-localization method using ego-motion signatures and motion correlation between observer and target wearable-camera videos,person-localization; social; privacy; retrieval,first-person-video
Left/Right Hand Segmentation,model,2016-07,CVIU 2016,watch,egocentric,2016,https://arxiv.org/abs/1607.06264,https://arxiv.org/abs/1607.06264,,,Classic first-person hand-segmentation approach that distinguishes left and right hands and handles hand-to-hand occlusion with temporal superpixels,hand-segmentation; hand-detection; gesture-recognition,first-person-video; hands
Unsupervised Important Objects,model,2016-11,ICCV 2017,watch,egocentric,2016,https://arxiv.org/abs/1611.05335,https://arxiv.org/abs/1611.05335,,,Unsupervised important-object detection in first-person images using cross-pathway supervision between visual semantics and spatial object proposals,object-discovery; object-detection; attention-modeling,first-person-video; first-person-images
EGO-TOPO Environment Affordances,model,2020-01,CVPR 2020,watch,egocentric,2020,https://arxiv.org/abs/2001.04583,https://arxiv.org/abs/2001.04583,,,Learns environment affordances from egocentric video by building topological maps of interaction opportunities in scenes,affordance; scene-understanding; navigation; action-recognition,egocentric-video; annotations
In the Eye of the Beholder,model,2020-05,IEEE TPAMI,watch,egocentric,2020,https://arxiv.org/abs/2006.00626,https://arxiv.org/abs/2006.00626,,,"Analyzes gaze and action in first-person video, studying how wearer attention supports egocentric action understanding",gaze; action-recognition; attention-modeling,egocentric-video; gaze; annotations
Hands in Egocentric Vision Survey,survey,2019-12,IEEE TPAMI,open,egocentric,2019,https://arxiv.org/abs/1912.10867,https://arxiv.org/abs/1912.10867,,,"Survey organizing hand-focused egocentric vision work around localization, interpretation, applications, and major hand-annotated datasets",survey; hand-object; hand-detection; hand-segmentation; gesture-recognition,egocentric-video; hands
R3M,model,2022-03,CoRL 2022,open,egocentric,2022,https://arxiv.org/abs/2203.12601,https://arxiv.org/abs/2203.12601,,,"Robot-manipulation visual representation pretrained on Ego4D human video with time-contrastive learning, video-language alignment, and sparse representation regularization",representation-learning; robot-manipulation; imitation-learning; egocentric-pretraining,egocentric-video; natural-language; robot-camera-video
VIP,model,2022-09,ICLR 2023,open,egocentric,2022,https://arxiv.org/abs/2210.00030,https://arxiv.org/abs/2210.00030,,,Value-Implicit Pre-Training learns a visual representation and dense reward signal from Ego4D human videos for few-shot robot manipulation,representation-learning; reward-modeling; robot-manipulation; egocentric-pretraining,egocentric-video; robot-camera-video
SpatialWorld,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.09669,https://arxiv.org/abs/2606.09669,,,"760 human-annotated real-world spatial tasks across eight simulation backends, evaluated under vision-only partial observability and egocentric evidence gathering",spatial-reasoning; embodied-ai; benchmark; interactive-assistance; vlm-evaluation,egocentric-video; text-instructions; annotations
MotionWAM,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.09215,https://arxiv.org/abs/2606.09215,,,Real-time humanoid world-action model driven by a single egocentric camera and unified whole-body motion tokens,world-modeling; humanoid-control; loco-manipulation; active-perception; vla,egocentric-video; robot-actions; humanoid-motion; latent-actions
OSCAR,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.04463,https://arxiv.org/abs/2606.04463,,,Omni-embodiment action-conditioned video world model trained through a pipeline spanning robotics and egocentric human datasets,world-modeling; video-generation; robot-learning; cross-embodiment-transfer; vla,egocentric-video; robot-camera-video; robot-actions; hand-skeleton; generated-video
Face-vs-Body HRI Tracking,dataset,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.03694,https://arxiv.org/abs/2606.03694,,,"Custom-annotated egocentric HRI tracking dataset from a social robot, comparing face and body tracking under occlusion and engagement changes",hri; tracking; engagement-detection; social; assistive-systems,egocentric-video; annotations; body-tracking
EgoForce Motion,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.13041,https://arxiv.org/abs/2605.13041,,,Online full-body motion reconstruction from sparse egocentric observations using diffusion forcing for long-term real-time applications,human-motion; body-pose; egocentric-3d-pose; ar-vr; real-time-understanding,egocentric-video; head-pose; body-motion; hand-pose
SABER,dataset,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.09613,https://arxiv.org/abs/2605.09613,,,100+ hours of in-store retail action capture with head-mounted egocentric footage and exocentric 360-degree scene video for VLA adaptation,vla; robot-learning; manipulation; action-understanding; data-collection,egocentric-video; exocentric-video; action-labels; language-instructions
T-REN,model,2026-04,arXiv,open,egocentric,2026,https://arxiv.org/abs/2604.18573,https://arxiv.org/abs/2604.18573,https://github.com/savya08/T-REN,,"Text-aligned region-token encoder that improves dense vision-language alignment while reducing long-video token counts, including reported gains on Ego4D video object localization",video-language; object-grounding; representation-learning; efficient-inference,video; egocentric-video; text
EgoViT Object SSL,model,2026-03,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2603.13912,https://arxiv.org/abs/2603.13912,,,"Self-supervised object representation learning from continuous, uncurated first-person video using proto-object discovery and stability objectives",representation-learning; object-discovery; object-recognition; egocentric-pretraining,first-person-video; egocentric-video; depth; optical-flow
V-JEPA 2.1,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.14482,https://arxiv.org/abs/2603.14482,,,"Dense self-supervised video representation model reporting strong Ego4D short-term object-interaction anticipation and EPIC-KITCHENS high-level action-anticipation results, plus robot-grasping transfer",representation-learning; world-modeling; action-anticipation; short-term-object-interaction-anticipation; robot-learning,video; egocentric-video; dense-features
CoMe-VLA,model,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.04600,https://arxiv.org/abs/2602.04600,,,"Cognitive and memory-aware VLA for active perception, learned from large-scale egocentric human data and non-Markovian sensing strategies",vla; active-perception; manipulation; robot-learning; memory-and-long-context,egocentric-video; robot-actions; memory; language-instructions
Central Vision SSL,model,2026-02,arXiv,open,egocentric,2026,https://arxiv.org/abs/2602.04462,https://arxiv.org/abs/2602.04462,https://github.com/t9s9/central-vision-ssl,,Time-contrastive self-supervised object-representation learning from Ego4D-derived gaze-centered crops that mimic central vision and temporal slowness in human visual experience,representation-learning; object-recognition; gaze; child-view-learning,egocentric-video; gaze; object-crops
RynnBrain,model,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.14979,https://arxiv.org/abs/2602.14979,,,"Open embodied foundation model family with variants for egocentric understanding, localization, physical reasoning, planning, navigation, and VLA",embodied-ai; egocentric-representation; spatial-reasoning; planning; vla,egocentric-video; video; text; robot-actions
UniT,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.19734,https://arxiv.org/abs/2604.19734,,,Unified latent action tokenizer that anchors massive egocentric human data to humanoid policy learning and world modeling,humanoid-control; cross-embodiment-transfer; world-modeling; imitation-learning; vla,egocentric-video; human-demonstrations; humanoid-motion; latent-actions
FunRec,model,2026-04,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2604.05621,https://arxiv.org/abs/2604.05621,,,"Functional 3D digital-twin reconstruction from in-the-wild egocentric RGB-D interaction videos, with articulated parts, kinematics, meshes, and URDF/USD export",3d-reconstruction; articulated-objects; scene-reconstruction; robot-learning; manipulation,egocentric-video; rgb-d; articulated-objects; object-meshes; digital-twins
METIS,model,2025-11,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2511.17366,https://arxiv.org/abs/2511.17366,,,"Dexterous VLA pretrained on EgoAtlas, a multi-source egocentric human and robotic data mixture unified under a consistent action space",vla; dexterous-manipulation; robot-learning; cross-embodiment-transfer; egocentric-pretraining,egocentric-video; robot-camera-video; robot-actions; hand-trajectories
Let Me Show You / RfV,model,2025-11,IROS 2025,watch,egocentric,2025,https://arxiv.org/abs/2511.05199,https://arxiv.org/abs/2511.05199,,,Retrieves relevant human egocentric demonstration clips and extracts affordance masks plus hand trajectories to condition robot manipulation policies,robot-learning; manipulation; retrieval; affordance; imitation-learning,egocentric-video; hand-trajectories; object-masks; language-instructions
EgoMI,model,2025-10,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2511.00153,https://arxiv.org/abs/2511.00153,,,"Egocentric Manipulation Interface for synchronized end-effector and active-head trajectories, retargeted to semi-humanoid robots",active-perception; whole-body-manipulation; imitation-learning; robot-learning; vla,egocentric-video; head-pose; hand-trajectories; robot-actions
Hand-VLA Pretraining,model,2025-10,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2510.21571,https://arxiv.org/abs/2510.21571,,,1M episodes and 26M frames converted from unscripted real-life egocentric hand videos into VLA-aligned action and language supervision,vla; robot-learning; egocentric-pretraining; action-segmentation; manipulation,egocentric-video; hand-motion; camera-motion; language-instructions
EMMA,model,2025-09,IEEE RA-L 2025,watch,egocentric,2025,https://arxiv.org/abs/2509.04443,https://arxiv.org/abs/2509.04443,,,Egocentric Mobile MAnipulation framework co-training human full-body motion data with static robot data for mobile manipulation policies,mobile-manipulation; robot-learning; imitation-learning; manipulation; cross-embodiment-transfer,egocentric-video; human-motion; robot-actions
TASTE-Rob,dataset,2025-03,CVPR 2025,watch,egocentric,2025,https://arxiv.org/abs/2503.11423,https://arxiv.org/abs/2503.11423,,,"100,856 task-oriented ego-centric hand-object interaction videos aligned with language instructions for video generation and robot imitation",video-generation; hand-object; robot-learning; imitation-learning; manipulation,egocentric-video; hand-object-interaction; language-instructions
Kaiwu,dataset,2025-03,IEEE RA-L 2025,watch,egocentric,2025,https://arxiv.org/abs/2503.05231,https://arxiv.org/abs/2503.05231,,,"11,664 synchronized assembling-action instances from 20 subjects with first-person video, gaze, hand motion, pressure, sound, EMG, mocap, robot, and environment streams",robot-learning; hri; industrial-assembly; sensor-fusion; manipulation,first-person-video; gaze; hand-motion; force; audio; emg; mocap; robot-actions
Humanoid-VLA,model,2025-02,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2502.14795,https://arxiv.org/abs/2502.14795,,,"Language, egocentric scene perception, and motion-control framework for universal humanoid control with video-conditioned fine-tuning",humanoid-control; vla; motion-generation; robot-learning; embodied-ai,egocentric-video; text-instructions; humanoid-motion
EgoSpeak,model,2025-02,NAACL 2025 Findings,watch,egocentric,2025,https://arxiv.org/abs/2502.14892,https://arxiv.org/abs/2502.14892,,,Real-time speech-initiation prediction for egocentric conversational agents over first-person RGB streaming video,streaming-assistant; multimodal-conversation; social; real-time-understanding; assistive-systems,egocentric-video; dialogue; speech
PRVQL,model,2025-02,ICCV 2025,watch,egocentric,2025,https://arxiv.org/abs/2502.07707,https://arxiv.org/abs/2502.07707,,,Progressive knowledge-guided refinement for egocentric visual query localization under appearance change and clutter,visual-query-localization; object-tracking; grounding-localization; episodic-memory,egocentric-video; object-crops; retrieval-features
Human Gaze Object-Centered Learning,model,2025-01,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2501.02966,https://arxiv.org/abs/2501.02966,,,Uses Ego4D-simulated months of egocentric experience and gaze-inspired central visual amplification to improve object-centered representation learning,representation-learning; gaze; object-recognition; egocentric-pretraining,egocentric-video; gaze; attention
R+X,model,2024-07,ICRA 2025,watch,egocentric,2024,https://arxiv.org/abs/2407.12957,https://arxiv.org/abs/2407.12957,,,"Retrieval-and-execution framework that teaches robots from long, unlabeled first-person videos of everyday human tasks",robot-learning; retrieval; imitation-learning; manipulation; cross-embodiment-transfer,first-person-video; language-instructions; robot-actions
Missing Modality Token / MMT,model,2024-01,CVPRW 2025,watch,egocentric,2024,https://arxiv.org/abs/2401.11470,https://arxiv.org/abs/2401.11470,,,"Missing-modality modeling for multimodal egocentric action recognition and moment localization on Ego4D, EPIC-KITCHENS, and EPIC-Sounds",multimodal-fusion; action-recognition; moment-query; robustness; egocentric-representation,egocentric-video; audio; imu; gaze
StarVLA,toolkit,2026-04,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2604.05014,https://arxiv.org/abs/2604.05014,,,Modular open-source VLA codebase supporting swappable VLM backbones such as Qwen-VL and world-model backbones such as Cosmos,tooling; vla; robot-learning; evaluation; qwen3-omni,robot-camera-video; robot-actions; language-instructions
HALOMI,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.18772,https://arxiv.org/abs/2606.18772,,,"Humanoid loco-manipulation from human demonstrations, extending UMI-style collection with egocentric head/wrist observations and head-hand trajectories",humanoid-control; manipulation; imitation-learning; active-perception; vla,egocentric-video; wrist-camera-video; human-demonstrations; robot-actions; humanoid-motion
HumanoidArena,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.17833,https://arxiv.org/abs/2606.17833,,,Benchmark for egocentric hierarchical whole-body learning with seven leg-critical humanoid-object and humanoid-scene interaction tasks,benchmark; humanoid-control; embodied-ai; manipulation; vla,egocentric-video; text-instructions; robot-actions; humanoid-motion
VLESA,benchmark,2026-06,arXiv,open,egocentric,2026,https://arxiv.org/abs/2606.03954,https://github.com/HanjiangHu/VLESA,https://github.com/HanjiangHu/VLESA,mit,Egocentric-video safety-assistance benchmark with goal-conditioned safety annotations for human activity monitoring,safety; activity-recognition; video-language; situated-assistance,egocentric-video; annotations; text
OVO-S-Bench,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.03890,https://arxiv.org/abs/2606.03890,,,"1,680 spatial-reasoning questions over 348 continuous egocentric videos with query timestamps and evidence intervals",spatial-reasoning; egocentric-video-qa; streaming-video-understanding; spatial-localization,egocentric-video; text; temporal-grounding; annotations
EgoPro-Bench,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.07299,https://arxiv.org/abs/2605.07299,,,"Personalized proactive-interaction benchmark with 2,400 evaluation videos, 12K+ training videos, and 12 domains",proactive-assistance; personalized-qa; streaming-video-understanding; intent-understanding,egocentric-video; text; annotations
Pro2Assist,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.04227,https://arxiv.org/abs/2605.04227,,,"Continuous step-aware proactive assistance with multimodal egocentric perception, including public-source data and an AR-glasses testbed",proactive-assistance; step-grounding; real-time-assistance; procedure-understanding,egocentric-video; smart-glasses; text; action-labels
EgoSPT / SPOT,dataset,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.20085,https://arxiv.org/abs/2605.20085,,,Egocentric spatially prompted manipulation trajectories with first-frame object/target grounding and 3D end-effector motion,trajectory-prediction; visual-grounding; manipulation; robot-learning; vla,egocentric-video; trajectories; object-masks; 3d-pose; text-instructions
EgoBabyVLM,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.19130,https://arxiv.org/abs/2605.19130,,,"Benchmarking cross-modal learning from naturalistic infant and adult egocentric video, including Machine-DevBench and an EgoBabyVLM challenge",child-view-learning; video-language; representation-learning; vlm-evaluation,child-view-video; egocentric-video; audio; language
EgoInteract,dataset,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.18214,https://arxiv.org/abs/2605.18214,,,Controllable egocentric-video simulator and synthetic dataset with dense spatial and temporal annotations,egocentric-simulation; temporal-segmentation; anticipation; next-active-object; ehoi,synthetic-egocentric; egocentric-video; annotations; scene-graphs
EggHand,model,2026-05,CVPR 2026 Findings,watch,egocentric,2026,https://arxiv.org/abs/2605.07642,https://jyoun9.github.io/EggHand/,,,Multimodal foundation model for egocentric hand-pose forecasting over EgoExo4D-style video-language and action signals,hand-forecasting; 3d-hand-pose; video-language; vla,egocentric-video; hand-pose; language; action-labels
Map-Mono-Ego,model,2026-05,ICIP 2026,watch,egocentric,2026,https://arxiv.org/abs/2605.20889,https://arxiv.org/abs/2605.20889,,,"Map-grounded global human-pose estimation from monocular egocentric video, with AIST-Living paired ego video and scanned environments",3d-human-pose; scene-understanding; localization; pose-and-body,egocentric-video; human-pose; scene-scans; camera-trajectory
Being-H0.7,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.00078,https://arxiv.org/abs/2605.00078,,,Latent world-action model learned from egocentric videos for future-aware reasoning and VLA policy learning,world-modeling; vla; robot-learning; future-state-prediction,egocentric-video; latent-actions; robot-actions
GazeVLA,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.22615,https://gazevla.github.io/,,,"Vision-language-action policy that pretrains on large-scale egocentric human data to capture gaze, intention, and action before robot fine-tuning",vla; gaze; intent-understanding; manipulation; robot-learning,egocentric-video; gaze; language; robot-actions
WARPED,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.10809,https://arxiv.org/abs/2604.10809,,,Wrist-aligned rendering pipeline that converts monocular egocentric human demonstrations into robot policy observations with 3D Gaussian Splatting,imitation-learning; cross-embodiment-transfer; robot-learning; manipulation,egocentric-video; wrist-camera-video; 3d-gaussian-splatting; hand-trajectories
PIE-V,benchmark,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.15134,https://arxiv.org/abs/2604.15134,,,Mistake-aware procedural egocentric-video benchmark injecting human-plausible mistakes and recovery corrections across Ego-Exo4D scenarios,mistake-detection; procedural-reasoning; action-and-procedure; procedural-assistance,egocentric-video; annotations; structured-reasoning
DP-DeGauss,model,2026-04,ICASSP 2026,watch,egocentric,2026,https://arxiv.org/abs/2604.07986,https://arxiv.org/abs/2604.07986,,,"Dynamic probabilistic Gaussian decomposition for egocentric 4D scene reconstruction, separating background, hands, and objects in first-person interaction",4d-reconstruction; scene-reconstruction; hand-object-interaction; neural-scene-reconstruction,egocentric-video; 3d-gaussian-splatting; hand-object-motion; 4d-scene
LifeEval,benchmark,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.00490,https://arxiv.org/abs/2603.00490,,,"4,075 high-quality QA pairs over continuous first-person streams for real-time task-oriented human-AI collaboration in daily life",egocentric-video-qa; interactive-assistance; real-time-assistance; multimodal-reasoning,egocentric-video; dialogue; text
SAVA-X,benchmark,2026-03,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2603.12764,https://arxiv.org/abs/2603.12764,,,"Ego-to-exo imitation-error detection over asynchronous, length-mismatched egocentric and exocentric videos, evaluated with EgoMe",mistake-detection; cross-view; imitation-learning; procedural-assistance,egocentric-video; exocentric-video; annotations
AG-EgoPose,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.25175,https://arxiv.org/abs/2603.25175,,,Attention-guided egocentric 3D human-pose estimation from fisheye camera input with dual motion/spatial streams,egocentric-3d-pose; 3d-human-pose; pose-and-body; motion,fisheye-video; 3d-human-pose; motion
Ego4D,dataset,2021-10,CVPR 2022,request,egocentric,2021,https://arxiv.org/abs/2110.07058,https://ego4d-data.org/,https://github.com/facebookresearch/Ego4d,Ego4D License Agreement,"3,670+ hours, 900+ camera wearers, 74 locations, 9 countries",episodic-memory; nlq; moment-query; hand-object; forecasting; social; audio-visual,video; audio; gaze; stereo; 3d; narrations
Ego-Exo4D,dataset,2023-11,CVPR 2024,request,egocentric,2023,https://arxiv.org/abs/2311.18259,https://ego-exo4d-data.org/,https://github.com/facebookresearch/projectaria_tools,Ego-Exo4D License Agreement,"1,286 hours, 740 participants, synchronized ego/exo views",ego-exo; skilled-activity; cross-view; proficiency; pose,video; audio; gaze; imu; 3d-point-clouds; camera-poses; language
EPIC-KITCHENS-100,dataset,2020-06,IJCV 2022,open,egocentric,2020,https://arxiv.org/abs/2006.13256,https://epic-kitchens.github.io/,https://github.com/epic-kitchens/epic-kitchens-100-annotations,CC BY-NC 4.0,"100 hours, 20M frames, 90K action segments, 45 kitchens",action-recognition; action-detection; anticipation; retrieval; domain-adaptation,video; audio; narrations; object-boxes
HD-EPIC,dataset,2025-02,CVPR 2025,open,egocentric,2025,https://arxiv.org/abs/2502.04144,https://arxiv.org/abs/2502.04144,,not specified,"41 hours, 9 kitchens, 59K fine-grained actions, 26K VQA questions",vqa; action-recognition; audio-events; object-motion; gaze-reasoning,video; audio; gaze; object-masks; digital-twins; 3d
HoloAssist,dataset,2023-09,ICCV 2023,open,egocentric,2023,https://arxiv.org/abs/2309.17024,https://holoassist.github.io/,,not specified,"166 hours, 350 instructor-performer pairs",interactive-assistance; mistake-detection; intervention-prediction; hand-forecasting,rgb; depth; gaze; hand; head; imu; audio
Assembly101,dataset,2022-03,CVPR 2022,open,egocentric,2022,https://arxiv.org/abs/2203.14712,https://assembly-101.github.io/,,not specified,"4,321 videos, 8 static plus 4 egocentric views, 1M fine-grained action segments",procedural-action; anticipation; temporal-segmentation; mistake-detection,multiview-video; egocentric-video; 3d-hand-pose
Ego-1K,dataset,2026-03,CVPR 2026,open,egocentric,2026,https://arxiv.org/abs/2603.13741,https://huggingface.co/datasets/facebook/ego-1k,,fair-noncommercial-research-license,"Nearly 1,000 synchronized multiview egocentric videos",dynamic-scene-understanding; novel-view-synthesis; 3d-reconstruction,multiview-video; calibration
EgoObjects,dataset,2023-09,ICCV 2023,open,egocentric,2023,https://arxiv.org/abs/2309.08816,https://github.com/facebookresearch/EgoObjects,,MIT,"9K+ videos, 650K annotations, 368 categories in pilot release",object-detection; instance-understanding; continual-learning,video; object-boxes; instance-labels
EgoBlind,dataset,2025-03,NeurIPS 2025 D&B,watch,egocentric,2025,https://arxiv.org/abs/2503.08221,https://arxiv.org/abs/2503.08221,,,"1,392 first-person videos from blind and visually impaired users; 5,311 questions",assistive-videoqa; blind-assistance; egocentric-vqa,first-person-video; qa
EgoDex,dataset,2025-05,ICLR 2026,watch,egocentric,2025,https://arxiv.org/abs/2505.11709,https://arxiv.org/abs/2505.11709,,,"829 hours, 194 tabletop tasks, Apple Vision Pro capture",dexterous-manipulation; imitation-learning; vla,egocentric-video; 3d-hand-pose; head-pose
EgoVerse,dataset,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.07607,https://egoverse.ai/,,,"1,362 hours, 80K episodes, 1,965 tasks, 2,087 demonstrators",robot-learning; vla; human-demonstrations,egocentric-video; annotations; tooling
OpenEgo,dataset,2025-09,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2509.05513,https://arxiv.org/abs/2509.05513,,,"1,107 hours across six public datasets, 290 manipulation tasks",dexterous-manipulation; standardized-pretraining; vla,egocentric-video; hand-pose; action-primitives
Dobb-E / HoNY,dataset,2023-11,arXiv,partial,egocentric,2023,https://arxiv.org/abs/2311.16098,https://dobb-e.com/,https://github.com/notmahi/dobb-e,,"Homes of New York dataset: 13 hours, 22 homes, 216 environments, 5,620 trajectories, 1.5M RGB-D frames collected with an iPhone-based Stick",robot-learning; imitation-learning; manipulation; data-collection,rgb; depth; robot-actions; wrist-camera-video
Universal Manipulation Interface / UMI,toolkit,2024-02,RSS 2024,open,egocentric,2024,https://arxiv.org/abs/2402.10329,https://umi-gripper.github.io/,https://github.com/real-stanford/universal_manipulation_interface,MIT,"Hand-held GoPro gripper interface and policy stack for in-the-wild robot teaching, including example and cup-arrangement data workflows",robot-learning; imitation-learning; cross-embodiment-transfer; bimanual-manipulation; data-collection,wrist-camera-video; imu; robot-actions
FastUMI,dataset,2024-09,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2409.19499,https://arxiv.org/abs/2409.19499,,,"UMI redesign with a reported 10,000+ real-world demonstration trajectories across 22 everyday manipulation tasks",robot-learning; imitation-learning; manipulation; data-collection,wrist-camera-video; robot-actions
FastUMI-100K,dataset,2025-10,arXiv,open,egocentric,2025,https://arxiv.org/abs/2510.08022,https://huggingface.co/datasets/IPEC-COMMUNITY/FastUMI_100k_lerobot,https://github.com/MrKeee/FastUMI-100K,apache-2.0,"100K+ UMI-style demonstration trajectories across 54 household manipulation tasks, standardized in LeRobot v2.1 with multi-view wrist-mounted fisheye images",robot-learning; imitation-learning; cross-embodiment-transfer; bimanual-manipulation; data-collection,wrist-camera-video; robot-actions; language-instructions
MV-UMI,toolkit,2025-09,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2509.18757,https://arxiv.org/abs/2509.18757,,,"Multi-view Universal Manipulation Interface adding third-person context to the egocentric wrist camera, evaluated across three manipulation tasks",robot-learning; imitation-learning; cross-embodiment-transfer; manipulation; data-collection,wrist-camera-video; third-person-video; robot-actions
UMIGen,model,2025-11,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2511.09302,https://arxiv.org/abs/2511.09302,,,Cloud-UMI handheld collection and visibility-aware point-cloud generation for egocentric 3D observations and cross-embodiment imitation learning,robot-learning; imitation-learning; cross-embodiment-transfer; 3d-reconstruction; data-collection,wrist-camera-video; point-clouds; robot-actions
Hoi!,dataset,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.04884,https://arxiv.org/abs/2512.04884,,,"3,048 force-grounded articulated-manipulation sequences over 381 objects and 38 environments, spanning human hand, wrist camera, UMI gripper, and Hoi gripper embodiments",robot-learning; imitation-learning; cross-embodiment-transfer; contact-understanding; hand-object,wrist-camera-video; robot-actions; contact; articulated-objects
YUBI,dataset,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.10244,https://arxiv.org/abs/2606.10244,,,"UMI-based bimanual dexterous manipulation stack with a reported 8,434 hours, 1.20M episodes, and 119 tasks",robot-learning; imitation-learning; cross-embodiment-transfer; bimanual-manipulation; dexterous-manipulation,wrist-camera-video; robot-actions; 6dof
Open X-Embodiment / RT-X,dataset,2023-10,ICRA 2024,open,adjacent,2023,https://arxiv.org/abs/2310.08864,https://robotics-transformer-x.github.io/,,cc-by-4.0,"1M+ real robot trajectories, 22 robot embodiments, 60 pooled robot datasets (robot-mounted and wrist cameras, not human first-person)",vla; robot-learning; cross-embodiment-transfer; imitation-learning; manipulation,robot-camera-video; wrist-camera-video; language-instructions; robot-actions; rlds
Qwen-RobotNav,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.18112,https://arxiv.org/abs/2606.18112,,,Qwen-branded scalable navigation model trained on 15.6M samples with controllable task modes and visual-history observation parameters,navigation; robot-learning; spatial-reasoning; embodied-reasoning,robot-camera-video; language-instructions; trajectories
Qwen-RobotWorld,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.17030,https://arxiv.org/abs/2606.17030,,,Language-conditioned embodied video world model with an 8.6M video-text EWK corpus over 20+ embodiments and 500+ action categories,world-modeling; video-generation; robot-learning; manipulation; navigation,video; language-instructions; robot-actions; synthetic-video
RhinoVLA,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.07383,https://arxiv.org/abs/2606.07383,https://github.com/HuixiAI/RhinoVLA,,"Deployment-oriented VLA model using a token-efficient Qwen3-VL backbone, continuous Action Expert, View Registry, and hardware-aware edge execution",vla; robot-learning; manipulation; edge-ai,robot-camera-video; robot-actions; language-instructions
OpenEAI-Platform,toolkit,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.03392,https://arxiv.org/abs/2606.03392,,,Open-source embodied-AI hardware/software platform with OpenEAI-Arm and OpenEAI-VLA built on Qwen3-VL-4B plus a Diffusion Transformer action head,vla; robot-learning; manipulation; tooling; data-collection,robot-camera-video; robot-actions; language-instructions
SAGE-Nav,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.25497,https://arxiv.org/abs/2606.25497,,,"Hierarchical object-goal navigation framework combining LLM planning, dynamic scene graphs, and real-time egocentric visual observations",navigation; spatial-reasoning; embodied-reasoning; robot-learning,robot-camera-video; scene-graphs; language-instructions
USS,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.25880,https://arescheah.github.io/uss-project-page/,,,"Unified spatial-semantic prompting framework for embodied visual tracking with text, point, box, and mask prompts decoded into egocentric waypoints",tracking; navigation; object-grounding; robot-learning,robot-camera-video; object-boxes; object-masks; language-instructions
RoboAtlas,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.26046,https://arxiv.org/abs/2606.26046,,,"Contextual Active SLAM with OpenRoboVox semantic mapping, global semantic-map reasoning, egocentric VLM reasoning, and Unitree Go2 deployment",navigation; spatial-reasoning; robot-learning; scene-understanding,robot-camera-video; 3d; scene-graphs
iCrowdNav,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.26047,https://arxiv.org/abs/2606.26047,,,"Intention-aware robot crowd navigation using egocentric visual observations, occupancy features, and human-pose interaction encoding",navigation; intent-understanding; trajectory-prediction; robot-learning,robot-camera-video; human-pose; trajectories
Cloak,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.22836,https://arxiv.org/abs/2606.22836,,,Zero-shot cross-embodiment VLA training recipe that masks the robot end-effector from wrist-camera observations to improve embodiment-agnostic visual reasoning,vla; robot-learning; cross-embodiment-transfer; manipulation,wrist-camera-video; robot-actions; language-instructions
OpenHLM,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.22174,https://openhlm-project.github.io/,,,"Open recipe for whole-body humanoid loco-manipulation with teleoperation, VLA model design, and heterogeneous co-training with HuMI data",humanoid-control; whole-body-manipulation; loco-manipulation; vla,robot-camera-video; robot-actions; humanoid-motion; language-instructions
UniviewVLA,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.21501,https://arxiv.org/abs/2606.21501,,,Multiview VLA with world modeling that infers future multiview scene observations from standard agent and wrist cameras for occlusion-aware robot manipulation,vla; world-modeling; robot-learning; manipulation,robot-camera-video; wrist-camera-video; robot-actions; language-instructions
LEXI-SG,model,2026-05,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2605.13741,https://ori-drs.github.io/lexisg-web/,,,Monocular RGB open-vocabulary 3D scene-graph mapping system validated on Habitat-Matterport and self-collected egocentric office sequences,navigation; spatial-reasoning; scene-understanding; 3d-understanding,robot-camera-video; scene-graphs; 3d; rgb
NavWAM,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.13494,https://arxiv.org/abs/2606.13494,,,Navigation world-action model that predicts future egocentric visual observations and action sequences for goal-conditioned visual navigation,navigation; world-modeling; robot-learning; spatial-reasoning,robot-camera-video; trajectories; goal-specification
IntentNav,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.08029,https://arxiv.org/abs/2606.08029,,,"Spatial-visual object-navigation policy learned from human demonstrations, using frontier probing and spatial memory for partially observed scenes",navigation; robot-learning; spatial-reasoning; imitation-learning,robot-camera-video; trajectories; demonstrations
AlloSpatial,toolkit,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.08952,https://arxiv.org/abs/2606.08952,,,Agentic spatial-reasoning harness that converts local egocentric observations into allocentric spatial representations for foundation-model reasoning,spatial-reasoning; embodied-ai; evaluation; navigation,egocentric-observations; spatial-memory; text
World-Language-Action Model,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.05979,https://arxiv.org/abs/2606.05979,,,"Unified embodied foundation model jointly predicting subtasks, subgoal images, and robot actions while learning from egocentric videos and robot trajectories",world-modeling; vla; robot-learning; planning,egocentric-video; robot-camera-video; language-instructions; robot-actions
ManiSplat,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.10645,https://arxiv.org/abs/2606.10645,,,Manipulation trajectory synthesis from monocular video using decoupled 3D Gaussian Splatting for dynamic interactive scenes,manipulation; robot-learning; trajectory-synthesis; 3d-understanding,monocular-video; 3d-gaussian-splatting; trajectories
DexFuture,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.05699,https://arxiv.org/abs/2606.05699,,,"Hierarchical future-state visuomotor targeting for bimanual dexterous tool use, using future visual targets for robot policy learning",dexterous-manipulation; robot-learning; world-modeling; bimanual-manipulation,robot-camera-video; robot-actions; future-state-targets
GRAIL,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.05160,https://arxiv.org/abs/2606.05160,,,Digital-twin pipeline generating humanoid loco-manipulation demonstrations from 3D assets and video priors,humanoid-control; robot-learning; manipulation; simulation,3d-assets; video-priors; humanoid-motion; robot-actions
DataLadder,toolkit,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.16776,https://arxiv.org/abs/2606.16776,,,"Simulation-enabled interconversion toolchain for moving between robot trials, demonstrations, synthetic data, and evaluation assets in an embodied-data pyramid",robot-learning; data-conversion; simulation; evaluation,simulation; robot-actions; synthetic-data; evaluation
EgoInfinity,toolkit,2026-06,arXiv,partial,adjacent,2026,https://arxiv.org/abs/2606.17385,https://arxiv.org/abs/2606.17385,,other,"Web-scale 4D hand-object interaction data engine; the public 46 GB preview contains 104 Action100M clips with retargeted trajectories for Franka, G1, Robonaut 2, and XLeRobot",hand-object-interaction; cross-embodiment-transfer; robot-learning; manipulation; data-collection,video; hand-trajectories; object-trajectories; robot-actions
EgoSAT,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.24422,https://leiyj23.github.io/EgoSAT/,,,Comprehensive benchmark for streaming egocentric interaction understanding and long-horizon reasoning across first-person clips,streaming-qa; interaction-understanding; vision-language; benchmark,egocentric-video; streaming-video; natural-language
LUCID,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.11628,https://lucid-robot.github.io/,,,Embodiment-agnostic intent model trained from unstructured human videos and paired with simulation-trained sensorimotor policies for dexterous robot skill acquisition,robot-learning; dexterous-manipulation; imitation-learning; cross-embodiment-transfer; manipulation,video; rgb-d; human-demonstrations; robot-actions
Dexterous Point Policy,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.10614,https://arxiv.org/abs/2606.10614,,,"Point-based dexterous manipulation policy that transfers from raw human demonstration videos using task-relevant 3D keypoints for hands and objects, without robot demonstrations",dexterous-manipulation; robot-learning; cross-embodiment-transfer; manipulation; vla,video; hand-pose; object-trajectories; robot-actions
CHORD,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2607.00033,https://nvidia-isaac.github.io/video_to_data/chord/,,,"Object-centric contact-wrench guidance for long-horizon dexterous RL, with a 4,739-task bimanual simulation benchmark; reports 82.12% average success over 1,831 tasks, 90.77% whole-body success, and real-hardware transfer",dexterous-manipulation; robot-learning; imitation-learning; contact-rich-manipulation; reinforcement-learning,human-demonstrations; force; contact; trajectories; robot-actions; simulation
TopoRetarget,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.16272,https://arxiv.org/abs/2606.16272,,,"Interaction-preserving retargeting framework that converts human hand-object demonstrations into dexterous robot-hand references using sparse interaction graphs, Laplacian deformation, contact preservation, and kinematic constraints",dexterous-manipulation; robot-learning; retargeting; contact-rich-manipulation; cross-embodiment-transfer,human-demonstrations; hand-pose; object-trajectories; contact; robot-actions
V2P-Manip,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.16436,https://arxiv.org/abs/2606.16436,,,"Pipeline for learning dexterous manipulation from monocular human demonstration videos through 3D asset acquisition, trajectory estimation, physical refinement, and dexterous policy learning",dexterous-manipulation; robot-learning; cross-embodiment-transfer; manipulation,monocular-video; trajectories; hand-pose; robot-actions
SeeTraceAct,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.02745,https://arxiv.org/abs/2606.02745,,,Demo-conditioned VLA framework with visibility-aware future end-effector trace prediction and RoboCasa-DC cross-embodiment demonstration episodes,vla; robot-learning; imitation-learning; cross-embodiment-transfer; manipulation,video; trajectories; human-demonstrations; robot-actions
What Matters When Cotraining Robot Manipulation Policies on Everyday Human Videos?,dataset,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.06627,https://arxiv.org/abs/2606.06627,,,"Study and 532-video, 28-hour everyday-human-video dataset with triangulated hand labels for cotraining robot manipulation policies under low robot-data regimes",robot-learning; manipulation; cross-embodiment-transfer; human-centric-policy-learning,video; hand-pose; human-demonstrations; robot-actions
LARA,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.07100,https://arxiv.org/abs/2606.07100,,,Latent Action Representation Alignment framework that jointly optimizes latent action models and VLAs so abundant unlabeled human videos can improve robot action learning,vla; robot-learning; manipulation; latent-action-learning; cross-embodiment-transfer,video; latent-actions; robot-actions; language-instructions
Video2Sim2Real,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.08828,https://arxiv.org/abs/2606.08828,,,Full-stack dexterous skill-acquisition pipeline that reconstructs a simulator-ready digital twin and robot/object motion priors from a single human manipulation video,dexterous-manipulation; robot-learning; simulation; cross-embodiment-transfer; manipulation,video; point-clouds; trajectories; robot-actions
HOWTransfer,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.10743,https://arxiv.org/abs/2606.10743,,,Hand-object open-world transfer method that localizes contact in human video demonstrations and retargets wrist trajectories into robot-executable motions,robot-learning; hand-object-interaction; cross-embodiment-transfer; manipulation; contact-understanding,video; hand-trajectories; object-trajectories; robot-actions
"Retrieve, Don't Retrain",model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.15631,https://arxiv.org/abs/2606.15631,,,Retrieval-augmented VLA adaptation method that adds new tasks at test time by appending pool-side demonstrations such as human-hand videos instead of retraining,vla; robot-learning; retrieval; cross-embodiment-transfer; manipulation; world-modeling,video; trajectories; human-demonstrations; robot-actions
SERF,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.12956,https://arxiv.org/abs/2606.12956,,,"Spatiotemporal environment and robot feature map for long-horizon mobile manipulation, updated online from egocentric observations and proprioceptive state",mobile-manipulation; robot-learning; spatial-reasoning; long-horizon-reasoning; manipulation,egocentric-observations; robot-actions; point-clouds; trajectories
ImageWAM,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.19531,https://arxiv.org/abs/2606.19531,,,World-action model that reframes robot future prediction as image editing instead of dense video generation for action prediction,world-modeling; robot-learning; vla; manipulation; video-generation,robot-camera-video; robot-actions; generated-video
NORA,model,2025-04,arXiv,watch,adjacent,2025,https://arxiv.org/abs/2504.19854,https://arxiv.org/abs/2504.19854,,,3B generalist VLA model using Qwen2.5-VL-3B as the visual-language backbone and 970K real-world robot demonstrations,vla; robot-learning; manipulation; efficient-inference,robot-camera-video; robot-actions; language-instructions
EgoLive,dataset,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.23570,https://arxiv.org/abs/2604.23570,,,Large-scale real-world task-oriented egocentric routines,robot-manipulation; real-world-routines,egocentric-video; multimodal-annotations
World In Your Hands,dataset,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.24310,https://arxiv.org/abs/2512.24310,,,"Reported 1,000+ hours of multimodal human-centric manipulation data",dexterous-manipulation; vla; human-centric-policy-learning,egocentric-video; multiview; depth; hand-pose; wrist-trajectory
InterVLA,dataset,2025-08,ICCV 2025,watch,egocentric,2025,https://arxiv.org/abs/2508.04681,https://arxiv.org/abs/2508.04681,,,"11.4 hours, 1.2M frames, 2 egocentric and 5 exocentric videos",vla; human-object-human-interaction; motion-estimation; interaction-prediction,egocentric-video; exocentric-video; human-motion; object-motion; verbal-commands
EgoPAT3D / EgoPAT3Dv2,dataset,2024-03,ICRA 2024,open,egocentric,2024,https://arxiv.org/abs/2403.05046,https://arxiv.org/abs/2403.05046,,not specified,Original EgoPAT3D reports 1M+ RGB-D and IMU frames; v2 expands 3D action-target prediction,3d-action-target-prediction; hri; anticipation; manipulation-safety,rgb; depth; imu; 3d-target-labels
HOT3D,dataset,2024-06,CVPR 2025,open,egocentric,2024,https://arxiv.org/abs/2406.09598,https://facebookresearch.github.io/hot3d/,,CC BY-NC-SA 4.0,"833+ minutes, 3.7M+ images, 19 subjects, 33 objects",3d-hand-object-tracking; ar-vr; pose,aria; quest3; gaze; point-clouds; 3d-hand-pose; object-pose; camera-pose
HOI4D,dataset,2022-03,CVPR 2022,open,egocentric,2022,https://arxiv.org/abs/2203.01577,https://hoi4d.github.io/,,CC BY-NC 4.0,"2.4M RGB-D egocentric frames, 4,000+ sequences, 800 objects",4d-hoi; pose-tracking; action-segmentation,rgbd; point-clouds; 3d-hand-pose; object-pose; segmentation
H2O,dataset,2021-04,ICCV 2021,open,egocentric,2021,https://arxiv.org/abs/2104.11181,https://arxiv.org/abs/2104.11181,,not specified,"Synchronized multi-view RGB-D benchmark with two-hand 3D pose, 6D object pose, interaction labels, camera poses, object meshes, and scene point clouds for first-person hand-object interaction recognition",first-person-interaction-recognition; two-hand-pose; hand-object-pose,multiview-rgbd; 3d-hand-pose; object-pose; camera-pose; object-meshes; scene-point-clouds
ARCTIC,dataset,2022-04,CVPR 2023,request,egocentric,2022,https://arxiv.org/abs/2204.13662,https://arctic.is.tue.mpg.de/,,,2.1M frames,bimanual-manipulation; reconstruction; interaction-field-estimation,video; hand-meshes; object-meshes; contact; articulated-objects
FPHA,dataset,2017-04,CVPR 2018,open,egocentric,2017,https://arxiv.org/abs/1704.02463,https://guiggh.github.io/publications/first-person-hands/,,not specified,"100K+ RGB-D frames, 45 action classes, 26 objects",hand-action-recognition; 3d-hand-pose,rgbd; 3d-hand-pose; object-pose
EgoHOS,dataset,2022-08,ECCV 2022,open,egocentric,2022,https://arxiv.org/abs/2208.03826,https://github.com/owenzlz/EgoHOS,,MIT,"11,243 egocentric images",hand-object-segmentation; contact-understanding,rgb; hand-object-contact-masks
Ego2Hands,dataset,2020-11,arXiv,open,egocentric,2020,https://arxiv.org/abs/2011.07252,https://arxiv.org/abs/2011.07252,,not specified,"Large-scale RGB egocentric two-hand segmentation/detection dataset with semi-automatic masks, compositing-based data generation, and a manually annotated evaluation set for unseen-domain generalization",two-hand-segmentation; hand-detection,rgb; synthetic-composited-masks
Ego2HandsPose,dataset,2022-06,WACV 2024,open,egocentric,2022,https://arxiv.org/abs/2206.04927,https://arxiv.org/abs/2206.04927,,not specified,Ego2Hands extension adding 3D hand-pose annotations and fitting/tracking tools for color-based two-hand 3D global pose estimation in unseen domains,two-hand-3d-pose,rgb; 3d-hand-pose
EgoBody,dataset,2021-12,ECCV 2022,open,egocentric,2021,https://arxiv.org/abs/2112.07642,https://sanweiliti.github.io/egobody/egobody.html,,not specified,"125 HoloLens2 interaction sequences with RGB, depth, gaze, head/hand tracking, scene-calibrated multi-Kinect ground truth, and SMPL-X body pose/shape for social-partner motion estimation",3d-human-pose; body-shape; motion,hololens2; rgb; depth; gaze; head-pose; hand-pose; body-pose
GIMO,dataset,2022-04,ECCV 2022,open,egocentric,2022,https://arxiv.org/abs/2204.09443,https://github.com/y-zheng18/GIMO,,not specified,"Gaze-informed human-motion dataset with egocentric views, eye gaze, inertial-motion body poses, and scene scans for intent-aware motion prediction in context",gaze-informed-motion-prediction; human-motion; scene-context,egocentric-video; gaze; body-pose; scene-scans
EgoHumans,benchmark,2023-05,ICCV 2023,partial,egocentric,2023,https://arxiv.org/abs/2305.16487,https://arxiv.org/abs/2305.16487,,,125K+ egocentric images,multi-human-tracking; 3d-pose; mesh-recovery,egocentric-video; multiview-video; 2d-pose; 3d-pose; mesh
UnrealEgo,dataset,2022-08,ECCV 2022,open,egocentric,2022,https://arxiv.org/abs/2208.01633,https://4dqv.mpi-inf.mpg.de/UnrealEgo/,,not specified,"Large-scale synthetic stereo-fisheye egocentric 3D human-pose corpus from virtual eyeglasses, covering diverse motions and naturalistic in-the-wild renderings",egocentric-3d-pose,synthetic-stereo; 3d-human-pose
xR-EgoPose,dataset,2019-07,GitHub,open,egocentric,2019,https://arxiv.org/abs/1907.10045,https://github.com/facebookresearch/xR-EgoPose,,custom,383K photorealistic synthetic HMD fisheye frames plus real-data evaluation for monocular egocentric 3D body-pose estimation under self-occlusion and perspective distortion,xr-pose-estimation,synthetic-egocentric; 3d-human-pose
EgoGTA / EgoPW-Scene,dataset,2022-12,CVPR 2023,partial,egocentric,2022,https://arxiv.org/abs/2212.11684,https://arxiv.org/abs/2212.11684,,,Synthetic scene-aware egocentric pose data supporting depth-inpainted scene constraints for monocular fisheye 3D human-pose estimation in occluded human-scene interactions,scene-aware-3d-human-pose; egocentric-depth; human-scene-interaction,synthetic-egocentric; scene-depth; human-pose
EgoTracks,benchmark,2023-01,NeurIPS 2023,open,egocentric,2023,https://arxiv.org/abs/2301.03213,https://arxiv.org/abs/2301.03213,,,"Ego4D-derived long-term object-tracking benchmark emphasizing large ego-motion, hand-object occlusion, target exits/re-entries, appearance changes, and re-detection",long-term-object-tracking; redetection,
TREK-150,benchmark,2021-08,project page,open,egocentric,2021,https://arxiv.org/abs/2108.13665,https://machinelearning.uniud.it/datasets/trek150/,,,150 densely annotated first-person tracking sequences from EPIC-KITCHENS for systematic evaluation of manipulated-object tracking under FPV motion and interaction,single-object-tracking,
Project Aria Datasets,collection,2023-08,arXiv,open,egocentric,2023,https://arxiv.org/abs/2308.13561,https://www.projectaria.com/datasets/,,,"Official Project Aria collection hub and tooling entry point for smart-glasses multimodal recordings, VRS/MPS processing, calibration, gaze, SLAM, and AR perception datasets",ar-perception; scene-understanding; wearable-sensing,aria; vrs; calibration; gaze; imu; mps
Aria Digital Twin,dataset,2023-06,ICCV 2023,open,egocentric,2023,https://arxiv.org/abs/2306.06362,https://www.projectaria.com/datasets/adt/,,CC BY-NC-SA 4.0,"200 sequences, 398 object instances",3d-machine-perception; object-tracking; scene-reconstruction,aria; rgb; monochrome; imu; 6dof; depth; segmentation; synthetic-rendering
Aria Everyday Activities,dataset,2024-02,arXiv,open,egocentric,2024,https://arxiv.org/abs/2402.13349,https://www.projectaria.com/datasets/aea/,,CC BY-NC-SA 4.0,143 daily activity sequences across five indoor locations,everyday-activity; neural-scene-reconstruction; prompted-segmentation,aria; trajectory; point-cloud; gaze; speech
Nymeria,dataset,2024-06,ECCV 2024,open,egocentric,2024,https://arxiv.org/abs/2406.09905,https://www.projectaria.com/datasets/nymeria/,,CC BY-NC-SA 4.0,"300 hours, 264 participants, 50 locations",motion-language; body-tracking; action-recognition,aria; eye-tracking; imu; body-motion; language; observer-view
Digital Twin Catalog,dataset,2025-04,CVPR 2025,open,egocentric,2025,https://arxiv.org/abs/2504.08541,https://www.projectaria.com/datasets/dtc/,,Project Aria DTC Dataset License Agreement,"2,000 scanned digital-twin objects",object-reconstruction; digital-twin-evaluation,3d-objects; dslr-images; aria-images
EgoLife,dataset,2025-03,CVPR 2025,partial,egocentric,2025,https://arxiv.org/abs/2503.03803,https://arxiv.org/abs/2503.03803,,,"300 hours, 6 participants, one week daily life",life-assistant; long-context-qa; memory,meta-aria; video; audio; third-person-reference; qa
Ego-EXTRA,dataset,2025-12,WACV 2026,open,egocentric,2025,https://arxiv.org/abs/2512.13238,https://fpv-iplab.github.io/Ego-EXTRA/,,not specified,"50 hours, 15K+ VQA pairs",expert-trainee-assistance; egocentric-vqa,video; dialogue; qa
EgoSchema,benchmark,2023-08,NeurIPS 2023,open,egocentric,2023,https://arxiv.org/abs/2308.09126,http://egoschema.github.io/,https://github.com/egoschema/EgoSchema,,"5K+ multiple-choice QA pairs, 250+ hours",long-video-qa; temporal-reasoning,
X-LeBench,benchmark,2025-01,EMNLP 2025,watch,egocentric,2025,https://arxiv.org/abs/2501.06835,https://arxiv.org/abs/2501.06835,,,"432 simulated life logs, 23 minutes to 16.4 hours",extremely-long-video-understanding; lifelogging; long-context-qa,
TeleEgo,benchmark,2025-10,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2510.23981,https://arxiv.org/abs/2510.23981,,,"3,291 human-verified QA items in streaming setting",streaming-assistant; memory; real-time-understanding; cross-memory-reasoning,egocentric-video; audio; text; timeline
EgoMemReason,benchmark,2026-05,COLM 2026,open,egocentric,2026,https://arxiv.org/abs/2605.09874,https://huggingface.co/datasets/Ted412/EgoMemReason,https://github.com/Ziyang412/EgoMemReason,cc-by-nc-4.0,"500 public multiple-choice questions over week-long EgoLife video across entity, event, and behavior memory; average 5.1 evidence segments and 25.9 hours of memory backtracking",long-horizon-memory; entity-memory; event-memory; behavior-memory; egocentric-video-reasoning,
EgoClip,dataset,2022-06,NeurIPS 2022,open,egocentric,2022,https://arxiv.org/abs/2206.01670,https://github.com/showlab/EgoVLP,,not specified,3.8M clip-text pairs,video-language-pretraining; retrieval,
RefEgo,benchmark,2023-08,ICCV 2023,open,egocentric,2023,https://arxiv.org/abs/2308.12035,https://github.com/shuheikurita/RefEgo,,,12K+ clips and 41 hours for video-based referring-expression comprehension,referring-expression-comprehension; object-grounding; referred-object-tracking,
EgoBench,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.27820,https://arxiv.org/abs/2605.27820,,,"1,045 egocentric-video-grounded interactive tasks",tool-using-agents; multimodal-reasoning; interaction,
EgoIntrospect,dataset,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.17262,https://ego-introspect.github.io/,,,"180 hours, 60 subjects",internal-state-reasoning; affect; intent; memory,video; audio; gaze; motion; physiological-signals; self-annotations
Minerva-Ego,benchmark,2026-05,arXiv,open,egocentric,2026,https://arxiv.org/abs/2605.15342,https://github.com/google-deepmind/neptune,,,"Complex egocentric visual-reasoning benchmark with multi-step multimodal questions, dense reasoning traces, and spatiotemporal object masks",multistep-reasoning; egocentric-video-qa; spatial-reasoning; visual-grounding; reasoning-traces,egocentric-video; text; segmentation; structured-reasoning
MA-EgoQA,benchmark,2026-03,arXiv,open,egocentric,2026,https://arxiv.org/abs/2603.09827,https://ma-egoqa.github.io/,,,1.7K questions over multiple egocentric streams,multi-agent-memory; social-reasoning; temporal-reasoning,
EASG-Bench,benchmark,2025-06,ICCV 2025 Workshop,open,egocentric,2025,https://arxiv.org/abs/2506.05787,https://github.com/fpv-iplab/EASG-bench,,,"Egocentric video QA benchmark generated from spatio-temporally grounded dynamic action scene graphs, exposing relation and temporal-ordering gaps in video-LLMs",scene-graph-qa; relation-reasoning; temporal-reasoning,
MyEgo,benchmark,2026-04,CVPR 2026,open,egocentric,2026,https://arxiv.org/abs/2604.01966,https://github.com/Ryougetsu3606/MyEgo,,,"541 long videos, 5K personalized questions",personalized-qa; ego-grounding; long-range-memory,
EgoToM,benchmark,2025-03,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2503.22152,https://arxiv.org/abs/2503.22152,,,"Ego4D-based theory-of-mind QA benchmark over camera-wearer goals, beliefs, and next actions, showing MLLMs trail humans on belief and future-action reasoning",theory-of-mind; goal-reasoning; belief-reasoning; next-action,
EgoSound,benchmark,2026-02,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2602.14122,https://arxiv.org/abs/2602.14122,,,"7,315 validated QA pairs across 900 videos",sound-understanding; audio-visual-reasoning; spatial-localization; causal-reasoning,
EgoEsportsQA,benchmark,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.12320,https://arxiv.org/abs/2604.12320,,,"1,745 QA pairs from professional matches across three first-person shooter games",first-person-qa; virtual-egocentric-reasoning; esports-perception; tactical-reasoning,first-person-video; qa
Causal-Plan-1M,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.01810,https://arxiv.org/abs/2606.01810,,,Million-scale explicit reasoning traces over egocentric videos,causal-reasoning; planning; next-state-estimation,
AssistQ,benchmark,2022-03,ECCV 2022,open,egocentric,2022,https://arxiv.org/abs/2203.04203,https://showlab.github.io/assistq/,,,531 QA samples from 100 instructional videos,task-assistance; affordance-centric-qa; action-planning,
EgoTaskQA,benchmark,2022-10,NeurIPS 2022,benchmark,egocentric,2022,https://arxiv.org/abs/2210.03929,https://arxiv.org/abs/2210.03929,,,"Egocentric task-understanding QA benchmark over action dependencies/effects, intents/goals, and agents' beliefs with descriptive, predictive, explanatory, and counterfactual questions",task-understanding; causal-reasoning; counterfactual-qa; belief-reasoning,
HowToDIV,dataset,2025-08,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2508.11192,https://arxiv.org/abs/2508.11192,,,"507 conversations, 6,636 QA pairs, 24 hours",procedural-task-assistance; instructional-dialogue; egocentric-qa,egocentric-video; dialogue; qa
EgoThink,benchmark,2023-11,CVPR 2024,benchmark,egocentric,2023,https://arxiv.org/abs/2311.15596,https://arxiv.org/abs/2311.15596,,,"First-person VQA benchmark spanning six capability groups and twelve dimensions, evaluating 18 VLMs and exposing persistent first-person perspective reasoning gaps",first-person-vqa; vlm-evaluation; embodied-reasoning,
VidEgoThink,benchmark,2024-10,ICLR 2025,benchmark,egocentric,2024,https://arxiv.org/abs/2410.11623,https://arxiv.org/abs/2410.11623,,,"Ego4D-derived embodied-video benchmark for video QA, hierarchy planning, visual grounding, and reward modeling, with GPT-4o-assisted generation and human filtering",video-qa; hierarchy-planning; visual-grounding; reward-modeling,
EgoCoT-Bench,benchmark,2026-05,arXiv,open,egocentric,2026,https://arxiv.org/abs/2605.19559,https://dstardust.github.io/EgoCoT/,,,"3,172 QA pairs over 351 egocentric videos",grounded-cot; operation-centric-reasoning; evidence-consistency,
NoRA,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.04806,https://arxiv.org/abs/2606.04806,,,"1,420 first-person video clips",normative-action-reasoning; grounded-reasonableness; safety,
Ego2Web,benchmark,2026-03,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2603.22529,https://arxiv.org/abs/2603.22529,,,Web-agent benchmark where agents must use egocentric video evidence about a physical scene to execute grounded online web tasks,web-agents; egocentric-grounding; physical-digital-assistance,
Pause and Think,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.00616,https://arxiv.org/abs/2606.00616,,,"Reasoning-centric proactive-assistance benchmark for grounded action suggestions over egocentric videos, emphasizing temporal consistency and when to pause before acting",assistive-action-suggestion; grounded-reasoning; temporal-consistency,
Ego-R1,benchmark,2025-06,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2506.13654,https://arxiv.org/abs/2506.13654,,,"Ego-CoTT-25K, Ego-QA-4.4K, and Ego-R1 Bench",ultra-long-video-qa; chain-of-tool-thought; temporal-retrieval,
EgoTrigger / HME-QA,benchmark,2025-08,ISMAR 2025 / TVCG,watch,egocentric,2025,https://arxiv.org/abs/2508.01915,https://arxiv.org/abs/2508.01915,,,340 human-annotated first-person QA pairs from full-length Ego4D videos curated for audio,memory-assistance; audio-triggered-capture; energy-efficient-smart-glasses; episodic-memory,first-person-video; audio; qa
Charades-Ego,dataset,2018-04,CVPR 2018,open,egocentric,2018,https://arxiv.org/abs/1804.09626,https://prior.allenai.org/projects/charades,,not specified,"68.8 hours, 68,536 activity instances",action-recognition; localization; captioning; domain-transfer,first-person-video; third-person-video; descriptions
MECCANO,dataset,2020-10,WACV 2021,open,egocentric,2020,https://arxiv.org/abs/2010.05654,https://iplab.dmi.unict.it/MECCANO/,,not specified,"Industrial-like motorbike assembly egocentric dataset with RGB/depth/gaze variants for egocentric hand-object interaction, active-object detection, anticipation, and action recognition",ehoi; active-object-detection; action-recognition; anticipation,rgb; depth; gaze
EgoProceL,dataset,2022-07,ECCV 2022,open,egocentric,2022,https://arxiv.org/abs/2207.10883,https://sid2697.github.io/egoprocel/,,not specified,"62 hours, 130 subjects, 16 tasks",procedure-learning; key-step-localization,
EGO-CH,dataset,2020-02,Pattern Recognition Letters 2020,partial,egocentric,2020,https://arxiv.org/abs/2002.00899,https://arxiv.org/abs/2002.00899,,,"27+ hours, 70 subjects, 2 cultural sites, 26 environments, 200+ points of interest",cultural-heritage; localization; object-recognition; retrieval; survey-prediction,egocentric-video; location-labels; poi-labels; surveys
EgoExoLearn,dataset,2024-03,CVPR 2024,open,egocentric,2024,https://arxiv.org/abs/2403.16182,https://github.com/OpenGVLab/EgoExoLearn,,MIT,120 hours,cross-view-association; planning; skill-assessment,egocentric-video; demonstration-video; gaze; annotations
KrishnaCam / OAK,dataset,2021-08,ICCV 2021,partial,egocentric,2021,https://arxiv.org/abs/2108.11005,https://oakdata.github.io/,,,"OAK: 80 snippets, about 17.5 hours, 105 object categories from KrishnaCam",continual-learning; object-detection; lifelogging,egocentric-video; object-boxes
EgoK360,dataset,2020-10,ICIP 2020,partial,egocentric,2020,https://arxiv.org/abs/2010.08055,https://egok360.github.io/,,,"First-person 360-degree activity-recognition dataset with activity and sub-action annotations, designed for spherical egocentric video understanding",360-video; activity-recognition,first-person-360-video; action-labels
EgoTraj,dataset,2026-05,arXiv,open,egocentric,2026,https://arxiv.org/abs/2605.19004,https://github.com/yehiahmad/EgoTraj,,custom,75 Meta Quest Pro navigation sequences,trajectory-prediction; navigation; assistive-systems,rgb; head-pose; gaze; scene-labels
VISOR,benchmark,2022-09,NeurIPS 2022,open,egocentric,2022,https://arxiv.org/abs/2209.13064,https://epic-kitchens.github.io/VISOR/,,cc-by-nc-4.0,"EPIC-KITCHENS pixel-level annotation benchmark with 272K manual masks, 257 object classes, 9.9M interpolated dense masks, 67K hand-object relations, and 36 hours from 179 untrimmed videos",hand-segmentation; active-object-segmentation; relations,
EPIC-Sounds,benchmark,2023-02,ICASSP 2023,open,egocentric,2023,https://arxiv.org/abs/2302.00646,https://epic-kitchens.github.io/epic-sounds/,,,EPIC-KITCHENS audio annotation benchmark with 78.4K categorized audible event/action segments across 44 classes plus 39.2K non-categorized segments and material labels for object-collision sounds,audio-event-recognition,
EPIC-Fields,benchmark,2023-06,NeurIPS 2023,open,egocentric,2023,https://arxiv.org/abs/2306.08731,https://epic-kitchens.github.io/epic-fields/,,,"3D camera and neural-rendering augmentation for EPIC-KITCHENS, reconstructing 96% of videos with 19M registered frames over 99 hours across 45 kitchens",3d-fields; spatial-reasoning,
EgoVLP,model,2022-06,NeurIPS 2022,open,egocentric,2022,https://arxiv.org/abs/2206.01670,https://github.com/showlab/EgoVLP,,,"Egocentric video-language pretraining framework introducing EgoClip (3.8M Ego4D clip-text pairs), EgoNCE, and EgoMCQ, transferring to EPIC-KITCHENS, Charades-Ego, and Ego4D tasks",video-language-pretraining; retrieval; ego4d-transfer,
EgoVLPv2,model,2023-07,ICCV 2023,open,egocentric,2023,https://arxiv.org/abs/2307.05463,https://shramanpramanick.github.io/EgoVLPv2/,,,"Second-generation egocentric VLP model with cross-modal fusion inside the video/language backbones, reducing downstream fusion overhead while improving multiple VL tasks",video-language-pretraining; cross-modal-fusion,
LaViLa,model,2022-12,CVPR 2023,open,egocentric,2022,https://arxiv.org/abs/2212.04501,https://arxiv.org/abs/2212.04501,,,"LLM-narrated video-language pretraining that densifies Ego4D-style narrations, improving EGTEA classification by 10.1 points and EPIC-KITCHENS-100 retrieval by 5.9 points",video-language-representation; narration-generation,
EgoNCE++,model,2024-05,ICLR 2025,open,egocentric,2024,https://arxiv.org/abs/2405.17719,https://github.com/xuboshen/EgoNCEpp,,,EgoHOIBench plus asymmetric contrastive EgoNCE++ objective with hard verb/noun negatives and object-centric text-to-video learning for fine-grained hand-object VLP,open-vocabulary-hoi; video-language-pretraining,
EgoDTM,model,2025-03,arXiv,open,egocentric,2025,https://arxiv.org/abs/2503.15470,https://github.com/xuboshen/EgoDTM,,,"Depth- and text-aware egocentric VLP model with pseudo-depth supervision, lightweight 3D-aware decoding, and hand-object-enriched captions for 3D-aware representation learning",3d-aware-vlp; depth-text-pretraining,
EgoVLM,model,2025-06,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2506.03097,https://arxiv.org/abs/2506.03097,,,Qwen2.5-VL-3B-based egocentric video model optimized with GRPO-style policy learning for first-person video reasoning,egocentric-video-reasoning; policy-optimization,egocentric-video; text; mllm
EgoGraph,model,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.23709,https://arxiv.org/abs/2602.23709,,,"Training-free temporal knowledge-graph framework for ultra-long egocentric videos, accumulating people/object/location/event relations for multi-day memory QA on EgoLifeQA and EgoR1-bench",temporal-knowledge-graph; ultra-long-video-qa,
GroundVQA,model,2023-12,CVPR 2024,open,egocentric,2023,https://arxiv.org/abs/2312.06505,https://github.com/Becomebright/GroundVQA,,,"Long egocentric video QA model unifying temporal query grounding and answering, using LLM-generated supervision and close-ended evaluation to improve QaEgo4D and Ego4D-NLQ",long-video-qa; temporal-grounding; ego4d-nlq,
HiERO,model,2025-05,ICCV 2025,watch,egocentric,2025,https://arxiv.org/abs/2505.12911,https://arxiv.org/abs/2505.12911,,,"Weakly supervised hierarchical activity-thread representation from narrated egocentric videos, improving EgoMCQ/EgoNLQ and zero-shot procedure learning (+12.5 F1 on EgoProceL)",hierarchical-activity-reasoning; egomcq; egonlq; procedure-learning,
EgoAgent,model,2025-02,ICCV 2025,watch,egocentric,2025,https://arxiv.org/abs/2502.05857,https://arxiv.org/abs/2502.05857,,,"Unified first-person agent transformer that interleaves states and actions to jointly learn representation, future-state prediction, and action prediction in egocentric worlds",joint-predictive-agent; future-state-prediction; action-prediction,
StillFast,model,2023-04,CVPRW 2023,open,egocentric,2023,https://arxiv.org/abs/2304.03959,https://iplab.dmi.unict.it/stillfast/,,,"End-to-end Ego4D short-term object-interaction anticipation model combining still image and video cues to localize next-active objects, predict verbs, and estimate contact timing",short-term-object-interaction-anticipation; next-active-object; ego4d,
CONE,model,2022-11,ECCV 2022 Workshop,watch,egocentric,2022,https://arxiv.org/abs/2211.08776,https://arxiv.org/abs/2211.08776,,,"Ego4D NLQ challenge coarse-to-fine temporal grounding pipeline with sliding-window prefiltering and EgoVLP-based moment ranking (blind [email protected] 15.26, [email protected] 9.24)",natural-language-query; video-language-grounding; ego4d-nlq,
EgoHandTrajPred / USST,model,2023-07,ICCV 2023,open,egocentric,2023,https://arxiv.org/abs/2307.08243,https://actionlab-cv.github.io/EgoHandTrajPred,,,"Egocentric 3D hand-trajectory forecasting framework with uncertainty-aware state-space Transformer, velocity constraints, visual prompt tuning, and H2O/EgoPAT3D evaluations",3d-hand-trajectory-forecasting; anticipation; h2o; egopat3d,
Project Aria Tools,toolkit,2023-08,arXiv,open,egocentric,2023,https://arxiv.org/abs/2308.13561,https://github.com/facebookresearch/projectaria_tools,,,"Official Project Aria SDK for reading VRS recordings, calibration, MPS/SLAM outputs, trajectories, gaze, and dataset artifacts across the Aria ecosystem",vrs-loading; calibration; mps; gaze; trajectory,
ActiveMimic,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.06194,https://arxiv.org/abs/2606.06194,,,Egocentric human-video pretraining framework that recovers synchronized hand-object and viewpoint dynamics from active perception,robot-learning; vla; egocentric-pretraining; active-perception; manipulation,egocentric-video; hand-object-motion; camera-motion
EgoPriMo,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.08495,https://arxiv.org/abs/2606.08495,,,Egocentric motion prior for interactive humanoid control learned from egocentric human demonstrations,humanoid-control; robot-learning; vla; motion-generation; imitation-learning,egocentric-video; human-demonstrations; humanoid-motion
FEEL,dataset,2026-03,arXiv,open,egocentric,2026,https://arxiv.org/abs/2603.15847,https://arxiv.org/abs/2603.15847,,cc-by-4.0,"Public 2 GB benchmark with 26 trials, 700 episodes, 39,497 frames, and 65.8 minutes of Aria RGB/gaze/hand/IMU aligned to Paxini grasp-force totals; the paper reports a larger roughly 3M-frame collection",physical-action-understanding; contact-understanding; hand-object; robot-learning,egocentric-video; force; glove-sensors; hand-object-contact
IMPACT,dataset,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.10409,https://arxiv.org/abs/2604.10409,,,"Synchronized five-view RGB-D industrial assembly dataset with ego-exo capture, bimanual annotations, state tracking, and anomaly-recovery supervision",industrial-assembly; procedural-action; anomaly-detection; ego-exo; skill-assessment,egocentric-video; exocentric-video; rgbd; bimanual-annotations; state-labels
MM-Conv,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.21796,https://arxiv.org/abs/2605.21796,,,"Referential-communication benchmark from 6.7 hours of egocentric VR interaction with synchronized speech, motion, gaze, and 3D scene geometry",referential-grounding; 3d-dialogue; multimodal-conversation; situated-assistance,egocentric-vr; speech; motion; gaze; 3d-scene
GBAT,toolkit,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.22962,https://arxiv.org/abs/2605.22962,,,AI-powered annotation toolkit for egocentric eye-tracking and video data in child-caregiver interaction studies,annotation-tooling; gaze-analysis; child-caregiver-interaction; wearable-sensing,egocentric-video; eye-tracking; gaze; annotations
OpenGlass Event Eyewear,toolkit,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.07431,https://arxiv.org/abs/2606.07431,,,Open-source low-power AI eyewear platform with event-based vision support for on-device smart-glasses research,ar-wearables; edge-ai; prototyping; on-device-perception,smart-glasses; event-camera; embedded-ml; wearable-sensors
Beyond Motion Primitives,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.27464,https://arxiv.org/abs/2605.27464,,,160K-sample Ego4D-derived dataset for behavioral activity recognition from head-mounted IMU,ar-wearables; behavioral-activity-recognition; imu; proactive-assistance,head-mounted-imu; egocentric-context-labels
EgoRelight,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.28401,https://arxiv.org/abs/2605.28401,,,Egocentric HMD-based human capture and illumination recovery for relightable photoreal avatar rendering,ar-vr; avatar-rendering; relighting; egocentric-human-capture,hmd-video; egocentric-capture; illumination; human-avatar
PAWS,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.25539,https://arxiv.org/abs/2603.25539,,,Extracts object articulations from large-scale hand-object interactions in egocentric videos,articulation-perception; 3d-scene-understanding; robot-learning; hand-object,egocentric-video; hand-object-interaction; articulated-objects
Static Scene Reconstruction from Dynamic Egocentric Videos,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.22450,https://arxiv.org/abs/2603.22450,,,Mask-aware reconstruction pipeline adapting static 3D reconstruction backbones to long-form dynamic egocentric video,scene-reconstruction; ar-vr; 3d-scene-understanding; dynamic-scenes,egocentric-video; 3d-reconstruction; masks
Ego-InBetween,model,2026-04,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2604.17749,https://arxiv.org/abs/2604.17749,,,Egocentric instructed visual state-transition generation for object transformations between initial and target states,world-modeling; state-transition-generation; action-modeling; hand-object,egocentric-video; text-instructions; generated-video
Syn2Seq Exo-to-Ego,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.13793,https://arxiv.org/abs/2604.13793,,,Exo-to-ego video generation via sequential interpolation from synchronized third-person views and camera poses,ego-exo; video-generation; view-synthesis; world-modeling,egocentric-video; exocentric-video; camera-pose; generated-video
VGGT-Segmentor,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.13596,https://arxiv.org/abs/2604.13596,,,Geometry-enhanced instance segmentation across disparate egocentric and exocentric views,cross-view-segmentation; ego-exo; remote-collaboration; embodied-ai,egocentric-video; exocentric-video; geometry; segmentation
Personal Point of View 3DGS,model,2026-04,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2604.23803,https://arxiv.org/abs/2604.23803,,,Evaluation of dynamic 3D Gaussian splatting methods for egocentric scene reconstruction under rapid camera motion and dynamic interactions,ar-vr; scene-reconstruction; 3d-understanding; dynamic-scenes,egocentric-video; 3d-gaussian-splatting; scene-reconstruction
EgoReasoner,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.06561,https://arxiv.org/abs/2603.06561,,,"Task-adaptive structured thinking for egocentric 4D reasoning over fixture interactions, viewpoint-relative locations, and object movement",4d-reasoning; spatial-reasoning; object-tracking; video-language,egocentric-video; 4d-scene; structured-reasoning
Gaze-Regularized VLMs,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.23190,https://arxiv.org/abs/2603.23190,,,Gaze-regularized VLM framework for ego-centric behavior understanding and intent/action focus,behavior-understanding; gaze-reasoning; action-recognition; intent-understanding,egocentric-video; gaze; vlm
Recurrent Reasoning VLM,model,2026-03,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2603.17312,https://arxiv.org/abs/2603.17312,,,Recurrent vision-language reasoning for estimating progress in long-horizon embodied tasks,task-progress-estimation; procedural-reasoning; embodied-ai; long-video-understanding,egocentric-video; video-language; task-progress
Ego4OOD,benchmark,2026-01,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2601.17056,https://arxiv.org/abs/2601.17056,,,Egocentric video domain-generalization benchmark built around covariate-shift scoring for action recognition,domain-generalization; action-recognition; robustness; covariate-shift,egocentric-video; action-labels; domain-shifts
Robust Egocentric Visual Attention,model,2026-01,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2601.01818,https://arxiv.org/abs/2601.01818,,,Language-guided scene-context-aware framework for robust egocentric visual attention prediction,visual-attention-prediction; gaze; scene-context; ar-wearables,egocentric-video; language; attention
HD-EPIC VQA T-CoT,model,2026-01,CVPR 2025 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2601.10228,https://arxiv.org/abs/2601.10228,,,"HD-EPIC VQA challenge solution with query/choice preprocessing, Qwen2.5-VL fine-tuning, temporal chain-of-thought prompting, and post-processing",hd-epic-vqa; long-video-reasoning; video-language; challenge-solution,egocentric-video; video-qa; mllm
Event-VStream,model,2026-01,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2601.15655,https://arxiv.org/abs/2601.15655,,,Event-aware real-time long-video stream understanding framework using semantically coherent event units,streaming-video-understanding; long-context-video; video-language; real-time-assistance,long-video; streaming-video; event-segments
V-Nutri,model,2026-04,CVPR 2026 MetaFood Workshop,watch,egocentric,2026,https://arxiv.org/abs/2604.11913,https://arxiv.org/abs/2604.11913,,,"Dish-level nutrition estimation from egocentric cooking videos, using preparation process evidence rather than final-dish images only",nutrition-estimation; cooking-understanding; procedural-reasoning; healthcare-assistance,egocentric-video; cooking-video; nutrition-labels
Trajectory-Conditioned Egocentric Prediction,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.20388,https://arxiv.org/abs/2605.20388,,,"Conditions egocentric future prediction on future camera trajectory to disambiguate actions, plans, and outcomes",action-anticipation; future-prediction; trajectory-conditioning; embodied-ai,egocentric-video; camera-trajectory; future-prediction
Zero-Shot Ego Object ReID,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.26383,https://arxiv.org/abs/2605.26383,,,Zero-shot object re-identification in egocentric kitchen videos via multi-stage SAM3 feature fusion,object-reidentification; kitchen-understanding; tracking; zero-shot,egocentric-video; object-crops; sam-features
Mamba Ego Action Recognition,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2605.24302,https://arxiv.org/abs/2605.24302,,,Cross-modal egocentric action-recognition model combining RGB and hand-skeleton streams via VideoMamba and CLS-token fusion,action-recognition; hand-object; multimodal-fusion; state-space-models,egocentric-video; hand-skeleton; rgb
Watch Remember Reason,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.07433,https://arxiv.org/abs/2606.07433,,,"Human-view video understanding framework for MLLMs organized around watching, remembering, and reasoning",long-video-understanding; memory; reasoning; mllm,human-view-video; multimodal-video; long-video
AnchorWorld,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.07326,https://arxiv.org/abs/2606.07326,,,Embodied egocentric world simulation with 3D-human-motion interaction integrity and view-based evolution customization,world-modeling; egocentric-simulation; video-generation; embodied-ai,egocentric-video; 3d-human-motion; generated-video
WristCompass,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.30671,https://arxiv.org/abs/2605.30671,,,Learns kinematic coupling as a visual concept for recovering ego-camera orientation in manipulation video,egocentric-calibration; imitation-learning; hand-object; robot-learning,egocentric-video; wrist-camera; hand-motion; camera-orientation
Objects Before Words,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.12985,https://arxiv.org/abs/2606.12985,,,Object-first contrastive learning bias for grounding caregiver language in child-view egocentric video,language-grounding; child-view-learning; egocentric-representation; weak-supervision,child-view-video; caregiver-speech; objects
Continual Child-View Learning,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.05115,https://arxiv.org/abs/2606.05115,,,Continual multimodal learning framework that processes a child's egocentric input stream in chronological order,continual-learning; language-grounding; child-view-learning; egocentric-representation,child-view-video; speech; chronological-stream
Understanding-Enhanced Ego Mistake Detection,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.02120,https://arxiv.org/abs/2606.02120,,,"Model-collaboration method for detecting incorrect user actions from egocentric video, using Qwen3-VL Embedding in the fine-grained reasoning branch",mistake-detection; procedural-assistance; action-reasoning; challenge-solution,egocentric-video; mllm; action-labels
VISTA,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2605.20901,https://arxiv.org/abs/2605.20901,,,V-JEPA Integrated StillFast Temporal Anticipator for Ego4D Short-Term Object Interaction Anticipation at EgoVis 2026,short-term-object-interaction-anticipation; ego4d; challenge-solution; hand-object,egocentric-video; object-boxes; action-labels
JFAA,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2605.20904,https://arxiv.org/abs/2605.20904,,,JEPA-based Future Action Anticipation method for the EPIC-KITCHENS-100 Action Anticipation Challenge at EgoVis 2026,action-anticipation; epic-kitchens; challenge-solution,egocentric-video; action-labels; v-jepa-features
FROST-STA,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2606.00694,https://arxiv.org/abs/2606.00694,,,Frozen dense-feature submission for Ego4D Short-Term Object Interaction Anticipation at EgoVis 2026,short-term-object-interaction-anticipation; ego4d; challenge-solution; hand-object,egocentric-video; object-boxes; dense-features
TAP-JEPA,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2606.00662,https://arxiv.org/abs/2606.00662,,,Runner-up EPIC-KITCHENS-100 Action Anticipation Challenge submission using frozen V-JEPA 2.1 future-latent probing and two-stage score fusion,action-anticipation; epic-kitchens; challenge-solution,egocentric-video; action-labels; v-jepa-features
HiERO-StepG,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2605.31227,https://arxiv.org/abs/2605.31227,,,Hierarchical activity-understanding solution for the Ego4D Step Grounding Challenge,step-grounding; procedure-understanding; ego4d; challenge-solution,egocentric-video; step-labels; hierarchical-activities
TempRet,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2605.24470,https://arxiv.org/abs/2605.24470,,,Temporal enhancement and two-stage reranking for the CVPR 2026 EPIC-KITCHENS-100 Multi-Instance Retrieval Challenge,video-text-retrieval; epic-kitchens; challenge-solution,egocentric-video; text; retrieval-features
CASTLE2026 Team WDL,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2606.00712,https://arxiv.org/abs/2606.00712,,,Evidence-aware multimodal reasoning pipeline for 600+ hours of multi-perspective egocentric CASTLE Challenge video,long-video-qa; castle; challenge-solution; memory-reasoning,egocentric-video; transcripts; auxiliary-images; multimodal-context
CuriosAI CASTLE,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2605.27800,https://arxiv.org/abs/2605.27800,,,"CASTLE Challenge submission using per-person timelines, speaker-resolved transcripts, multi-VLM captions, and search-verify-answer reasoning",long-video-qa; castle; challenge-solution; memory-reasoning,egocentric-video; transcripts; captions; multimodal-context
MARS CASTLE,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2605.18176,https://arxiv.org/abs/2605.18176,,,"Multimodal agentic reasoning with source selection for CASTLE Challenge questions over multi-day, multi-perspective egocentric recordings",long-video-qa; castle; challenge-solution; memory-reasoning,egocentric-video; exocentric-video; transcripts; gaze; thermal; physiological-signals
EgoCross Domain-Wise Inference,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2606.00829,https://arxiv.org/abs/2606.00829,,,Nearly training-free domain-wise inference strategy for EgoCross source-limited egocentric video QA,egocentric-video-qa; domain-shift; challenge-solution; inference-strategy,egocentric-video; qa; support-set
HD-EPIC Semantic-Visual Evidence,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2605.29402,https://arxiv.org/abs/2605.29402,,,HD-EPIC VQA Challenge solution that separates semantic and visual evidence for efficient long-video reasoning,hd-epic-vqa; long-video-reasoning; challenge-solution,egocentric-video; video-qa; semantic-evidence; visual-evidence
UniversalVTG,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.08522,https://arxiv.org/abs/2604.08522,,,"Lightweight cross-dataset foundation model for video temporal grounding, surfaced in the Ego4D/egocentric grounding scan",video-temporal-grounding; long-video-understanding; video-language; transfer-learning,video; text; temporal-grounding
EgoCom / Ego audio-visual correspondence,dataset,2023-07,CVPR 2024,open,egocentric,2023,https://arxiv.org/abs/2307.04760,http://vision.cs.utexas.edu/projects/ego_av_corr/,,not specified,Egocentric video with spatial audio for conversation and audio-visual correspondence tasks,foundation-video; video-language,
EasyCom,dataset,2021-07,arXiv,open,egocentric,2021,https://arxiv.org/abs/2107.04174,https://arxiv.org/abs/2107.04174,,not specified,AR glasses egocentric multi-channel audio and wide-FOV RGB for noisy conversations,foundation-video; video-language,
Look and Tell,dataset,2025-10,NeurIPS 2025 Workshop,watch,egocentric,2025,https://arxiv.org/abs/2510.22672,https://arxiv.org/abs/2510.22672,,,"25 participants, Project Aria plus stationary cameras, gaze/speech/video, 3D reconstructions",foundation-video; video-language,
EgoVLA,model,2025-07,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2507.12440,https://rchalyang.github.io/EgoVLA/,,,VLA training from egocentric human videos plus robot fine-tuning,robot-learning; manipulation; vla,
EgoEngine,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.12604,https://egoengine.github.io/,,,"Framework converting egocentric RGB human manipulation videos into temporally aligned robot observation videos and feasible executable robot action trajectories, with zero-shot dexterous policy learning from human videos",robot-learning; manipulation; vla,
EgoAERO,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.08057,https://arxiv.org/abs/2606.08057,,,Asset-free conversion from a single egocentric RGB-D demonstration; introduces EgoDex-R in the paper,robot-learning; manipulation; vla,
HRDexDB,dataset,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.14944,https://arxiv.org/abs/2604.14944,,,"1.4K human/robot grasping trials, tactile, multiview video, egocentric video streams",robot-learning; manipulation; vla,
UnrealEgo2 / UnrealEgo-RW,dataset,2024-01,CVPR 2024,watch,egocentric,2024,https://arxiv.org/abs/2401.00889,https://arxiv.org/abs/2401.00889,,,Expanded stereo egocentric pose datasets from synthetic and real-world capture,hand-object; 3d; pose,
TouchMoment,dataset,2026-04,CVPR 2026 Findings,watch,egocentric,2026,https://arxiv.org/abs/2604.12343,https://arxiv.org/abs/2604.12343,,,"4,021 egocentric videos, 8,456 annotated hand-object contact moments",hand-object; 3d; pose,
EgoFun3D,dataset,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.11038,https://arxiv.org/abs/2604.11038,,,"271 egocentric videos with 3D geometry, part segmentation, articulation and function-template annotations",hand-object; 3d; pose,
EgoEMG,dataset,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.05712,https://arxiv.org/abs/2605.05712,,,"41 participants, bilateral EMG, IMU, RGB, external RGB-D, mocap hand labels",hand-object; 3d; pose,
EgoEVHands,dataset,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.12297,https://github.com/ZJUWang01/EgoEV-HandPose,,,"Stereo event-camera egocentric hand dataset with 5,419 sequences and 3D/2D keypoints",hand-object; 3d; pose,
EventEgoHands,benchmark,2025-05,ICIP 2025,watch,egocentric,2025,https://arxiv.org/abs/2505.19169,https://arxiv.org/abs/2505.19169,,,Event-based egocentric 3D hand mesh reconstruction benchmark over N-HOT3D,hand-object; 3d; pose,
A multimodal RGB/events FPV hand dataset,dataset,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.10790,https://arxiv.org/abs/2606.10790,,,Synthetic event-based first-person hand detection from EgoHands plus v2e,hand-object; 3d; pose,
EgoExoMem,dataset,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.18734,https://arxiv.org/abs/2605.18734,,,2.6K MCQs across synchronized ego-exo videos,memory; qa; assistant,
EgoSelf,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.19564,https://abie-e.github.io/egoself_project/,,,Personalized egocentric assistant framework with graph memory,memory; qa; assistant,
EgoCross,dataset,2025-08,AAAI 2026,watch,egocentric,2025,https://arxiv.org/abs/2508.10729,https://github.com/MyUniverse0726/EgoCross,,,"About 1,000 QA pairs over surgery, industry, extreme sports, and animal-perspective clips",memory; qa; assistant,
Wrist-mounted ADL,dataset,2015-11,CVPR 2016,open,egocentric,2015,https://arxiv.org/abs/1511.06783,https://arxiv.org/abs/1511.06783,,not specified,Synchronized head and wrist wearable-camera daily activities,action-recognition; procedure,
Visual Experience Dataset / VEDB,dataset,2024-02,Journal of Vision,partial,egocentric,2024,https://arxiv.org/abs/2404.18934,http://tamaraberg.com/visualexperience/,,,240+ hours egocentric video with gaze/head tracking in classic literature,action-recognition; procedure,
FT-HID,dataset,2022-09,GitHub,open,egocentric,2022,https://arxiv.org/abs/2209.10155,https://github.com/ENDLICHERE/FT-HID,,not specified,90K+ RGB-D first- and third-person human interaction samples from 109 subjects,action-recognition; procedure,
LSC-ADL,dataset,2025-04,ACM MM 2025,open,egocentric,2025,https://arxiv.org/abs/2504.02060,https://arxiv.org/abs/2504.02060,,not specified,ADL annotations over lifelogging data generated with clustering plus human review,action-recognition; procedure,
SEED4D,dataset,2024-12,WACV 2025,open,adjacent,2024,https://arxiv.org/abs/2412.00730,https://seed4d.github.io/,,,"Synthetic ego-exo dynamic 4D generator and autonomous-driving dataset (16.8M images, vehicle inward/outward cameras, LiDAR; WACV 2025)",driving; 4d-reconstruction; multi-view,
HD-EPIC VQA Challenge,benchmark,2025-02,CVPR 2025,benchmark,egocentric,2025,https://arxiv.org/abs/2502.04144,https://arxiv.org/abs/2502.04144,,,"Derived benchmark suite built on HD-EPIC; tasks: Recipe, ingredient, nutrition, fine-grained action, 3D perception, object motion, gaze",benchmark; evaluation,
EgoEnv,model,2022-07,NeurIPS 2023 Oral,open,egocentric,2022,https://arxiv.org/abs/2207.11365,https://vision.cs.utexas.edu/projects/ego-env/,,,"Human-centric environment representation learning that links egocentric video to the camera wearer's local surroundings, trained in simulated 3D environments and transferred to HouseTours and Ego4D",video-language; representation-learning; scene-understanding; natural-language-query,
EgoMAS,model,2026-03,arXiv,open,egocentric,2026,https://arxiv.org/abs/2603.09827,https://ma-egoqa.github.io/,,,Shared-memory baseline for multi-agent egocentric video QA,video-language; representation-learning,
Ego-Exo representation transfer,model,2021-04,CVPR 2021,open,egocentric,2021,https://arxiv.org/abs/2104.07905,https://arxiv.org/abs/2104.07905,,,Distillation from third-person video using ego-specific latent signals,action; tracking; pose; hoi,
EgoPoseFormer,model,2024-03,ECCV 2024,open,egocentric,2024,https://arxiv.org/abs/2403.18080,https://github.com/ChenhongyiYang/egoposeformer,,,Transformer baseline for stereo egocentric 3D human pose estimation,action; tracking; pose; hoi,
EgoSTARK,model,2023-01,NeurIPS 2023,open,egocentric,2023,https://arxiv.org/abs/2301.03213,https://arxiv.org/abs/2301.03213,,,Adapted long-term tracker baseline for EgoTracks,action; tracking; pose; hoi,
EgoHOS model,model,2022-08,GitHub,open,egocentric,2022,https://arxiv.org/abs/2208.03826,https://github.com/owenzlz/EgoHOS,,,Context-aware hand-object segmentation and augmentation pipeline,action; tracking; pose; hoi,
AV-CONV,model,2023-12,CVPR 2024,open,egocentric,2023,https://arxiv.org/abs/2312.12870,https://vjwq.github.io/AV-CONV/,,,Audio-visual conversational graph prediction from ego/exo conversation,action; tracking; pose; hoi,
EgoAction,model,2026-05,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2605.24496,https://arxiv.org/abs/2605.24496,,,CVPR 2026 EPIC-KITCHENS action detection challenge pipeline,action; tracking; pose; hoi,
EgoAdapt,model,2026-05,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2605.24500,https://arxiv.org/abs/2605.24500,,,CVPR 2026 HD-EPIC VQA challenge inference-time adaptation pipeline,action; tracking; pose; hoi,
EgoObjects API,toolkit,2023-09,ICCV 2023,open,egocentric,2023,https://arxiv.org/abs/2309.08816,https://github.com/facebookresearch/EgoObjects,,,Working with category and instance-level egocentric object labels.,tooling,
EgoVid-5M,dataset,2024-11,NeurIPS 2025,open,egocentric,2024,https://arxiv.org/abs/2411.08380,https://egovid.github.io/,,not specified,5M curated egocentric clips at 1080p with fine-grained kinematic and high-level textual action annotations (NeurIPS 2025),video-generation; world-modeling; action-conditioned-generation,
AoE,dataset,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.23893,https://arxiv.org/abs/2602.23893,,,Always-on egocentric human video collection pipeline and corpus for embodied AI data scaling,robot-learning; data-collection; embodied-ai,
Seeing Across Views (MV-RoboBench),benchmark,2025-10,ICLR 2026,open,adjacent,2025,https://arxiv.org/abs/2510.19400,https://github.com/microsoft/MV-RoboBench,,,MV-RoboBench: 1.7K curated QA items over eight subtasks testing multi-view spatial reasoning of VLMs in robotic manipulation scenes (ICLR 2026),multi-view; robot-manipulation; vlm-evaluation,
EgoBrain,dataset,2025-06,ICLR 2026,request,egocentric,2025,https://arxiv.org/abs/2506.01353,https://huggingface.co/datasets/ut-vision/EgoBrain,,cc-by-nc-4.0,"Complete 1.6 TB release from 40 participants with synchronized 4K GoPro egocentric video, EEG, IMU, interval markers, and surveys for human action understanding",eeg; action-understanding; multimodal,egocentric-video; eeg; imu; natural-language; annotations
MM-Ego,model,2024-10,ICLR 2025,open,egocentric,2024,https://arxiv.org/abs/2410.07177,https://arxiv.org/abs/2410.07177,,,"Egocentric multimodal LLM with Memory Pointer Prompting; 7M-sample QA data engine and the EgoMemoria benchmark (629 videos, 7,026 questions)",video-language; memory; qa,
EgoStream,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.31557,https://arxiv.org/abs/2605.31557,,,"2,250 questions across seven memory dimensions with Answer Validity Windows, expanded to 8,528 recall-conditioned evaluations over streams up to 45.3 hours",streaming-memory; episodic-memory; qa,
Gesture-Based Egocentric Video QA,benchmark,2026-03,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2603.12533,https://arxiv.org/abs/2603.12533,,,Egocentric video QA grounded in the camera wearer's pointing and deictic gestures,gesture-grounding; qa; referential,
ExAct,benchmark,2025-06,arXiv,open,egocentric,2025,https://arxiv.org/abs/2506.06277,https://arxiv.org/abs/2506.06277,,,Video-language benchmark for expert action analysis and feedback over skilled egocentric/exocentric activity,expert-feedback; skill-assessment; video-language,
Home Action Genome / HOMAGE,dataset,2021-05,CVPR 2021,open,egocentric,2021,https://arxiv.org/abs/2105.05226,https://homeactiongenome.org/,,not specified,"27 participants, multi-modal synchronized ego and third-person views with 12 sensor types and hierarchical activity/action labels in home settings",action-recognition; multi-view; compositional,
EgoExo-Fitness,dataset,2024-06,ECCV 2024,open,egocentric,2024,https://arxiv.org/abs/2406.08877,https://github.com/iSEE-Laboratory/EgoExo-Fitness,,Apache-2.0,"Synchronized ego and exo fitness videos with keypoint verification, execution comments, and action quality scores (ECCV 2024)",ego-exo; action-quality; skill-assessment,
WEAR,dataset,2023-04,IMWUT 2024,open,egocentric,2023,https://arxiv.org/abs/2304.05088,https://mariusbock.github.io/wear/,,not specified,"22 participants, 18 outdoor workout activities, synchronized egocentric video and 3D acceleration across 11 locations (IMWUT 2024)",action-recognition; sensor-fusion; imu,
Egocentric HOI Detection,benchmark,2025-06,Expert Systems with Applications,open,egocentric,2025,https://arxiv.org/abs/2506.14189,https://arxiv.org/abs/2506.14189,,,New benchmark and method for detecting hand-object interactions in egocentric video,hand-object; detection; interaction,
SHOW3D,dataset,2026-03,CVPR 2026,open,egocentric,2026,https://arxiv.org/abs/2603.28760,https://arxiv.org/abs/2603.28760,,cc-by-nc-4.0,"2,137 in-the-wild scenes and about 20 hours from 38 participants, captured with eight exocentric cameras plus two Quest 3 egocentric views; 4.28M synchronized multiview frames, 8.55M ego images, 3D hand/object pose, shape, and text",hand-object; 3d-hand-object-tracking; 3d-reconstruction,multiview-video; 3d-hand-pose; object-pose; text
EgoTouch / TouchAnything,dataset,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.13083,https://jianyi2004.github.io/TouchAnything-Website/,,,"EgoTouch: 302 manipulation tasks across 4,530 episodes with synchronized egocentric and dual wrist-camera video, bimanual 3D hand pose (42 joints), and dense tactile pressure maps; TouchAnything is the accompanying tactile-estimation framework",hand-object; contact-understanding; bimanual-manipulation,egocentric-video; wrist-camera-video; 3d-hand-pose; contact
EgoEverything,benchmark,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.08342,https://arxiv.org/abs/2604.08342,,,"5,000+ MCQ QA pairs over 100+ hours of egocentric video, with questions generated using human attention/gaze signals for AR long-context understanding",long-context; video-qa; gaze-reasoning,egocentric-video; gaze; qa
EgoNight,benchmark,2025-10,ICLR 2026,open,egocentric,2025,https://arxiv.org/abs/2510.06218,https://insait-institute.github.io/EgoNight/,https://github.com/dehezhang2/EgoNight,,"Nighttime egocentric benchmark: EgoNight-VQA (3,658 QA over 90 videos, 12 QA types) plus day-night correspondence retrieval and egocentric depth estimation, with day-night aligned synthetic and real video",egocentric-vqa; egocentric-depth; retrieval,egocentric-video; qa; depth; synthetic-rendering
EgoIntent,benchmark,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.12147,https://arxiv.org/abs/2603.12147,,,"3,014 steps across 15 indoor/outdoor daily-life scenarios, evaluating local intent (What), global intent (Why), and next-step plan (Next), with future-frame leakage removed",intent; next-action; anticipation,egocentric-video; qa
EgoMAGIC,dataset,2026-04,arXiv,open,egocentric,2026,https://arxiv.org/abs/2604.22036,https://arxiv.org/abs/2604.22036,,CC BY 4.0,"3,355 egocentric field-medicine videos over 50 medical tasks from a head-mounted stereo camera with audio; 1.95M labels for 124 medical objects and an action-detection challenge",action-detection; object-detection; procedural-task-assistance,egocentric-video; stereo; audio; object-boxes
Audio Hallucination in Egocentric Video,benchmark,2026-04,ICASSP 2026,watch,egocentric,2026,https://arxiv.org/abs/2604.23860,https://arxiv.org/abs/2604.23860,,,"300 egocentric videos and 1,000 sound-focused questions probing audio hallucination in audio-visual LLMs that infer sounds from visual cues",audio-visual-reasoning; sound-understanding; evaluation,egocentric-video; audio; qa
Ego2World,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.13335,https://arxiv.org/abs/2605.13335,,,Compiles egocentric cooking videos (built on HD-EPIC) into executable symbolic worlds with graph-transition rules for belief-state planning under partial observability,planning; belief-reasoning; embodied-reasoning,egocentric-video; annotations
EgoCrowds / CrowdEraser,dataset,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.29036,https://arxiv.org/abs/2603.29036,,,EgoCrowds: semi-synthetic paired crowded/empty clips from real egocentric walking-tour footage; CrowdEraser is a video-diffusion framework that removes crowds to produce humanless environment walkthroughs,video-generation; scene-understanding; scene-reconstruction,egocentric-video; first-person-video
EgoPressure,dataset,2024-09,CVPR 2025,open,egocentric,2025,https://arxiv.org/abs/2409.02224,https://yiming-zhao.github.io/EgoPressure/,,not specified,"5.0 hours, 21 participants, a moving egocentric camera plus 7 stationary RGB-D cameras, with hand pose meshes and fine-grained per-contact touch pressure (CVPR 2025)",hand-object; contact-understanding; 3d-hand-pose,egocentric-video; rgbd; 3d-hand-pose; contact
EgoTL,benchmark,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.09535,https://arxiv.org/abs/2604.09535,,,Think-aloud (say-before-act) egocentric capture with word-level spoken reasoning and metric-scale spatial annotations over 100+ household tasks; benchmarks VLMs and world models on six task dimensions,long-context; planning; embodied-reasoning,egocentric-video; speech; annotations
SuperMemory-VQA,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.00825,https://arxiv.org/abs/2606.00825,,,"52.9 hours of AI-glasses daily activities with RGB, audio transcription, gaze, IMU, and SLAM; 4,853 human-verified QA pairs across object/location memory, intent recall, and scene recall",episodic-memory; long-context; qa,egocentric-video; audio; gaze; imu; point-clouds
EgoAVU,benchmark,2026-02,CVPR 2026,open,egocentric,2026,https://arxiv.org/abs/2602.06139,https://github.com/facebookresearch/EgoAVU,https://github.com/facebookresearch/EgoAVU,,Scalable audio-visual data engine producing EgoAVU-Instruct (3M QAs) and the human-verified EgoAVU-Bench (3K QAs) for egocentric audio-visual understanding (CVPR 2026 highlight),audio-visual-reasoning; sound-understanding; evaluation,egocentric-video; audio; qa
EgoScale,dataset,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.16710,https://arxiv.org/abs/2602.16710,,,"20,854 hours of action-labeled egocentric human video for VLA pretraining with a human-to-robot two-stage transfer recipe; reveals a log-linear human-data scaling law",vla; robot-learning; dexterous-manipulation,egocentric-video; action-labels
StableHand,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.18553,https://huajian-zeng.github.io/projects/stablehand/,,,"Quality-aware flow-matching baseline for world-space dual-hand motion estimation from egocentric video, evaluated on HOT3D and ARCTIC",3d-hand-pose; pose; hand-object,egocentric-video; 3d-hand-pose
EgoVideo,model,2024-06,CVPR 2024,open,egocentric,2024,https://arxiv.org/abs/2406.18070,https://github.com/OpenGVLab/EgoVideo,https://github.com/OpenGVLab/EgoVideo,,Egocentric video foundation model with a slow-fast adaptation recipe; winner of multiple Ego4D / EPIC-KITCHENS challenge tracks (CVPR 2024 EgoVis),foundation-video; video-language; representation-learning,egocentric-video; text
EgoHOD,model,2025-03,ICLR 2025,open,egocentric,2025,https://arxiv.org/abs/2503.00986,https://github.com/InternRobotics/EgoHOD,https://github.com/InternRobotics/EgoHOD,,Fine-grained hand-object-dynamics pretraining (HOD pipeline) for egocentric video representation learning; SOTA on EK-100 retrieval/classification and EGTEA (ICLR 2025),video-language-pretraining; hand-object; representation-learning,egocentric-video; hand; text
EgoM2P,model,2025-06,ICCV 2025,watch,egocentric,2025,https://arxiv.org/abs/2506.07886,https://arxiv.org/abs/2506.07886,,,"Egocentric multimodal, multitask pretraining framework over RGB, depth, gaze, and camera pose with a masked modeling objective",multimodal; representation-learning; standardized-pretraining,egocentric-video; depth; gaze; camera-pose
AlanaVLM,model,2024-06,EMNLP 2024,watch,egocentric,2024,https://arxiv.org/abs/2406.13807,https://arxiv.org/abs/2406.13807,,,Multimodal embodied-AI foundation model trained for egocentric video understanding with an egocentric video instruction dataset,video-language; embodied-ai; video-qa,egocentric-video; text
Vinci,model,2024-12,IMWUT 2025,open,egocentric,2025,https://arxiv.org/abs/2412.21080,https://github.com/OpenGVLab/vinci,https://github.com/OpenGVLab/vinci,,"Real-time, always-on embodied smart assistant built on an egocentric vision-language model for portable devices, with historical-context QA, task planning, and step-by-step visual demonstrations",real-time-understanding; streaming-assistant; task-assistance,egocentric-video; audio; text
AMEGO,model,2024-09,ECCV 2024,open,egocentric,2024,https://arxiv.org/abs/2409.10917,https://gabrielegoletto.github.io/AMEGO/,https://github.com/gabrielegoletto/AMEGO,,"Active-memory representation from a single long egocentric video, capturing key locations and object interactions for fast multi-query answering; introduces the Active Memories Benchmark (20K+ queries over EPIC-KITCHENS, ECCV 2024)",long-context; memory; episodic-memory,egocentric-video
EgoInstructor,model,2024-01,CVPR 2024,open,egocentric,2024,https://arxiv.org/abs/2401.00789,https://jazzcharles.github.io/Egoinstructor/,,,"Retrieval-augmented egocentric video captioning that retrieves relevant third-person instructional clips, trained with an EgoExoNCE cross-view objective (CVPR 2024)",captioning; video-language; cross-view,egocentric-video; exocentric-video; text
GroundNLQ,model,2023-06,CVPR 2023,open,egocentric,2023,https://arxiv.org/abs/2306.15255,https://github.com/houzhijian/GroundNLQ,https://github.com/houzhijian/GroundNLQ,,Two-stage pretrained multi-scale grounding model for Ego4D Natural Language Queries; champion of the CVPR 2023 Ego4D NLQ challenge,natural-language-query; temporal-grounding; ego4d,egocentric-video; text
OSGNet,model,2025-05,CVPR 2025,open,egocentric,2025,https://arxiv.org/abs/2505.04270,https://github.com/Yisen-Feng/OSGNet,https://github.com/Yisen-Feng/OSGNet,,Object-shot enhanced grounding network for egocentric video temporal grounding; top entry on Ego4D NLQ / Goal-Step moment queries (CVPR 2025),natural-language-query; temporal-grounding; object-grounding,egocentric-video; text
EgoThinker,model,2025-10,NeurIPS 2025,watch,egocentric,2025,https://arxiv.org/abs/2510.23569,https://arxiv.org/abs/2510.23569,,,Egocentric reasoning model with spatio-temporal chain-of-thought supervision and reinforcement fine-tuning for first-person video understanding,egocentric-video-reasoning; grounded-cot; video-qa,egocentric-video; text
Walk through Paintings,model,2026-01,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2601.15284,https://arxiv.org/abs/2601.15284,,,"Egocentric world models built from internet priors, generating first-person walkthroughs of static scenes such as paintings",world-modeling; video-generation; scene-understanding,egocentric-video
Temporal-Aware Ego VLM,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.27184,https://arxiv.org/abs/2603.27184,,,"Training scheme that incentivizes temporal awareness in egocentric video-understanding models, improving order- and timing-sensitive reasoning",temporal-reasoning; video-language; representation-learning,egocentric-video; text
EgoMotion,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.19105,https://arxiv.org/abs/2604.19105,,,Hierarchical reasoning plus diffusion framework for egocentric vision-language motion generation from first-person context,motion; motion-estimation; video-language,egocentric-video; motion; text
Gaze-SoM HOI Anticipation,model,2026-04,ICPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2604.03667,https://arxiv.org/abs/2604.03667,,,Leverages gaze and set-of-mark prompting in vision-language LLMs for hand-object-interaction anticipation from egocentric video,anticipation; hand-object; gaze-reasoning,egocentric-video; gaze; text
Sanpo-D,benchmark,2026-01,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2601.18100,https://arxiv.org/abs/2601.18100,,,"Fine-grained re-annotation of the Sanpo egocentric navigation data with spatial signals, benchmarking spatial-conditioned reasoning of VLMs over long first-person videos",spatial-reasoning; navigation; video-qa,egocentric-video; qa
EgoSurgery,dataset,2025-03,Healthcare Technology Letters,open,egocentric,2025,https://arxiv.org/abs/2503.18755,https://github.com/Fujiry0/EgoSurgery,https://github.com/Fujiry0/EgoSurgery,custom,Egocentric open-surgery video dataset family: EgoSurgery-Phase for surgical phase recognition and EgoSurgery-HTS for pixel-wise hand-tool segmentation of 14 surgical tools (MICCAI 2024),action-recognition; hand-object-segmentation; procedural-task-assistance,egocentric-video; segmentation; object-boxes
EgoEMS,dataset,2025-11,AAAI 2026,watch,egocentric,2025,https://arxiv.org/abs/2511.09894,https://arxiv.org/abs/2511.09894,,,"High-fidelity multimodal egocentric dataset for cognitive assistance in emergency medical services, capturing time-critical team actions in high-stakes scenarios",procedural-task-assistance; action-recognition; real-time-understanding,egocentric-video; audio; multimodal-annotations
LEMMA,dataset,2020-07,ECCV 2020,open,egocentric,2020,https://arxiv.org/abs/2007.15781,https://arxiv.org/abs/2007.15781,,not specified,"Multi-view multi-agent multi-task daily-activity dataset across 14 kitchens and living rooms, densely annotated with atomic actions and human-object interactions (ECCV 2020)",compositional; action-recognition; anticipation,egocentric-video; third-person-video; multimodal-annotations
EgoMe,dataset,2025-01,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2501.19061,https://arxiv.org/abs/2501.19061,,,"Real-world dataset and challenge for following a demonstrator via egocentric view, pairing exocentric demonstrations with egocentric imitation across everyday tasks",imitation-learning; cross-view; action-understanding,egocentric-video; exocentric-video
ENIGMA-360,dataset,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.09741,https://arxiv.org/abs/2603.09741,,,Ego-exo dataset for human behavior understanding in industrial scenarios with 360-degree exocentric and egocentric capture,action-understanding; ego-exo; procedure,egocentric-video; exocentric-video; first-person-360-video
EgoIllusion,benchmark,2025-08,EMNLP 2025,watch,egocentric,2025,https://arxiv.org/abs/2508.12687,https://arxiv.org/abs/2508.12687,,,"Benchmark for hallucinations in egocentric video understanding, probing where multimodal models fabricate objects, actions, and sounds",evaluation; multimodal-reasoning; audio-visual-reasoning,egocentric-video; qa
EgoExo-Con,benchmark,2025-10,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2510.26113,https://arxiv.org/abs/2510.26113,,,Synchronized ego-exo video pairs with human-refined queries for view-invariant temporal verification and grounding; introduces the View-GRPO training framework,cross-view; temporal-grounding; evaluation,egocentric-video; exocentric-video; qa
WearVQA,benchmark,2025-11,NeurIPS 2025,watch,egocentric,2025,https://arxiv.org/abs/2511.22154,https://arxiv.org/abs/2511.22154,,,"First wearable-device VQA benchmark for smart glasses: 2,520 image-question-answer triplets across 7 image domains and 10 cognitive task types under occluded, low-light, and blurry egocentric capture (NeurIPS 2025)",vqa; evaluation; wearable-sensing,egocentric-video; qa
WearVox,benchmark,2026-01,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2601.02391,https://arxiv.org/abs/2601.02391,,,"Egocentric multichannel voice-assistant benchmark for wearables, evaluating spoken interaction grounded in first-person audio-visual context",audio-visual-reasoning; streaming-assistant; evaluation,egocentric-video; audio; speech
SuperGlasses,benchmark,2026-02,CVPR 2026 Findings,watch,egocentric,2026,https://arxiv.org/abs/2602.22683,https://arxiv.org/abs/2602.22683,,,Benchmark evaluating vision-language models as intelligent agents for AI smart glasses across realistic wearable assistant tasks,tool-using-agents; evaluation; task-assistance,egocentric-video; qa
PhysBrain,model,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.16793,https://arxiv.org/abs/2512.16793,,,Uses human egocentric data as a bridge to adapt vision-language models toward physical intelligence and embodied control,embodied-ai; representation-learning; robot-learning,egocentric-video; text
EgoCampus,dataset,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.07668,https://arxiv.org/abs/2512.07668,,,Egocentric pedestrian eye-gaze dataset and model over campus walking routes with synchronized first-person video and gaze,gaze; navigation; trajectory-prediction,egocentric-video; gaze; scene-labels
Egocentric Clinical Intent,benchmark,2026-01,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2601.06750,https://arxiv.org/abs/2601.06750,,,Benchmark for egocentric clinical intent understanding by medical multimodal LLMs over first-person clinical procedure video,intent; procedural-task-assistance; evaluation,egocentric-video; qa
1M-HUGs / HUG,dataset,2026-06,arXiv,open,egocentric,2026,https://arxiv.org/abs/2606.17054,https://grasping.io/,,not specified,"Egocentric human-grasp dataset with 1M frames (27.8 hours), 6,707 object instances across 41 buildings, plus HUG-Bench with 90 unseen objects for zero-shot grasping",robot-learning; dexterous-manipulation; hand-object; manipulation,egocentric-video; rgbd; hand-pose; object-meshes
EgoKit,toolkit,2026-05,arXiv,open,egocentric,2026,https://arxiv.org/abs/2605.16797,https://egokit.chuange.org/,,,"Unified low-cost egocentric recording workflow across Android, iPhone, iPad, smart-glasses, and XR hosts, with synchronized ego and wrist views plus OpenXR hand tracking where available",data-collection; robot-learning; wearable-sensing; tooling,egocentric-video; wrist-camera-video; head-pose; hand-pose
MobileEgo Anywhere,toolkit,2026-05,arXiv,request,egocentric,2026,https://arxiv.org/abs/2605.05945,https://arxiv.org/abs/2605.05945,,cc-by-nc-4.0,"STERA-10M release with 200 hours, 584 smartphone sessions, 10M frames, 75,857 atomic actions, 9,922 episodes, 2,212 subgoals, and about 1.6 TB of RGB, depth, pose, IMU, MANO hands, mesh, and language",data-collection; robot-learning; vla; long-horizon-memory,egocentric-video; trajectory; action-labels
UCS-Bench / DirectMe,benchmark,2026-06,ICML 2026,open,egocentric,2026,https://arxiv.org/abs/2606.15200,https://github.com/cocowy1/UCS-Bench,https://github.com/cocowy1/UCS-Bench,,170+ hours of egocentric observations with 8.1K+ timestamped questions for user-centric continual spatial intelligence; DirectMe maintains streaming spatial memory,spatial-reasoning; episodic-memory; long-context; streaming-memory; qa,egocentric-video; qa; trajectory
StreamMemBench,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.14571,https://arxiv.org/abs/2606.14571,,,"Streaming benchmark over EgoLife egocentric streams, pairing each evidence anchor with an initial task and a future follow-up to test recall, feedback incorporation, and reuse",streaming-memory; memory-assistance; evaluation; egocentric-vqa,egocentric-video; qa; dialogue
Ego-METAS,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.02246,https://maria-sanvil.github.io/Ego-METAS-website/,,,"Online multimodal energy-efficient temporal action segmentation benchmark over 100+ hours from EgoExo4D, CMU-MMAC, and CaptainCook4D with RGB, audio, gaze, IMU, and monochrome streams",temporal-segmentation; action-segmentation; sensor-fusion; evaluation,egocentric-video; audio; gaze; imu; monochrome
Plan Watch Recover / EgoProactive,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.04970,https://arxiv.org/abs/2606.04970,,,"Proactive procedural-assistance benchmark introducing EgoProactive with out-of-plan annotations and Pro2Bench over Ego4D, EPIC-KITCHENS, EgoExo4D, HoloAssist, and HowTo100M",procedural-task-assistance; mistake-detection; planning; interactive-assistance,egocentric-video; qa; action-labels
EgoExoBench,benchmark,2025-07,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2507.18342,https://arxiv.org/abs/2507.18342,,,"First-/third-person video reasoning benchmark with 7,300+ QA pairs across 11 subtasks covering semantic alignment, viewpoint association, and temporal reasoning",ego-exo; cross-view; temporal-reasoning; video-qa; evaluation,egocentric-video; exocentric-video; qa
EgoProx,benchmark,2026-05,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2605.24456,https://arxiv.org/abs/2605.24456,,,"Egocentric 3D proximity reasoning benchmark organized across intention, exploration, exploitation, and chain-of-actions reasoning for MLLMs",spatial-reasoning; 3d; vqa; evaluation,egocentric-video; qa; 3d
BARISTA,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.12074,https://arxiv.org/abs/2605.12074,,,"185 egocentric coffee-preparation videos with per-frame scene graphs, persistent object IDs, masks, tracks, boxes, attributes, relations, hand-object interactions, activities, and process steps",scene-graph-qa; procedural-task-assistance; hand-object; temporal-reasoning,egocentric-video; object-masks; object-boxes; action-labels
TAVIS,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.07943,https://arxiv.org/abs/2605.07943,,,"Active-vision imitation-learning benchmark with TAVIS-Head (5 tasks), TAVIS-Hands (3 tasks), GR1T2 and Reachy2 embodiments, and the GALT anticipatory-gaze metric",imitation-learning; gaze; robot-learning; evaluation,egocentric-video; gaze; robot-actions
EgoPoint-Bench,benchmark,2026-04,ACL 2026,watch,egocentric,2026,https://arxiv.org/abs/2604.21461,https://arxiv.org/abs/2604.21461,,,11K+ simulated and real egocentric pointing samples spanning five evaluation dimensions and three levels of referential complexity,gesture-grounding; referential; spatial-reasoning; vqa,egocentric-video; qa; hand-pose
ReFocus / EM-QnF,benchmark,2026-04,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2604.24893,https://nsubedi11.github.io/refocus,,,"Episodic Memory with Questions and Feedback task for ambiguous egocentric NLQ, with feedback datasets and a plug-and-play Feedback Alignment Module",episodic-memory; natural-language-query; interactive-assistance; temporal-grounding,egocentric-video; qa; dialogue
HumanEgo,model,2026-05,arXiv,open,egocentric,2026,https://arxiv.org/abs/2605.24934,https://humanego-ai.github.io,https://github.com/TX-Leo/HumanEgo,cc-by-nc-4.0,"Zero-shot robot learning from minutes of egocentric video; the public 235.6 GB package contains 122 Aria recordings across two tasks, raw VRS/MPS data, segmentation, hand/object tracks, and 3D targets",robot-learning; imitation-learning; cross-embodiment-transfer; hand-object,egocentric-video; hand-pose; robot-actions
EgoGuide,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.14665,https://silicx.github.io/EgoGuide,,,"Robot-free demonstration collection with synchronized wrist and head/egocentric observations, online visual-geometric data-quality guidance, and a gated egocentric residual policy",data-collection; robot-learning; imitation-learning; vla,egocentric-video; wrist-camera-video; robot-actions
Ego-Pi,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.08107,https://egopipaper.github.io/,,,VLA fine-tuning study over egocentric human and robot data using the pi0.5 foundation model and dexterous five-finger embodiments,vla; robot-learning; cross-embodiment-transfer; dexterous-manipulation,egocentric-video; robot-actions; hand-pose
EgoTactile,model,2026-06,ICML 2026 spotlight,watch,egocentric,2026,https://arxiv.org/abs/2606.09243,https://egotactile.github.io/,,,Benchmark and diffusion/baseline models pairing egocentric video with full-hand grasp-pressure supervision for everyday-object interactions,contact-understanding; hand-object; robot-learning; evaluation,egocentric-video; hand-pose; contact; depth
EgoPressDiff,model,2026-06,ICASSP 2026,watch,egocentric,2026,https://arxiv.org/abs/2606.06872,https://egopressdiff.github.io/,,,"Conditional video diffusion model for UV-domain egocentric hand-pressure maps, conditioned on hand pose, 3D mesh vertices, and depth",contact-understanding; hand-object; robot-learning,egocentric-video; hand-pose; mesh; depth
EgoForce Hand Pose,model,2026-05,SIGGRAPH 2026,open,egocentric,2026,https://arxiv.org/abs/2605.12498,https://dfki-av.github.io/EgoForce,https://dfki-av.github.io/EgoForce,,"Monocular egocentric 3D hand-pose and shape reconstruction across fisheye, perspective, and wide-FOV camera models, with source code, data, and demo available",3d-hand-pose; hand-object; ar-vr; pose,egocentric-video; hand-pose; hand-meshes
PROSE,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.16569,https://rckola.github.io/prose/,,,Training-free RGB-only egocentric scene registration using VLM-derived object-level 3D scene graphs for persistent spatial memory in robots and AR,scene-reconstruction; spatial-reasoning; ar-perception; 3d,egocentric-video; 3d; object-masks
OSGNet + MLLM Reranking,model,2026-05,CVPR 2026 EgoVis,open,egocentric,2026,https://arxiv.org/abs/2605.20818,https://github.com/iLearn-Lab/CVPR25-OSGNet,https://github.com/iLearn-Lab/CVPR25-OSGNet,,"Champion solution for the Ego4D Episodic Memory Challenge 2026 Natural Language Queries and GoalStep tracks, reranking OSGNet candidates with an MLLM",natural-language-query; temporal-grounding; ego4d; video-language,egocentric-video; text
OmniEgo-R2,model,2026-05,CVPR 2026 EgoVis,open,egocentric,2026,https://arxiv.org/abs/2605.24481,https://github.com/Lee-zixu/OmniEgo-R2,https://github.com/Lee-zixu/OmniEgo-R2,,Routed reasoning framework for EgoCross; second place in both Source-Limited and Open-Source tracks of the 1st Cross-Domain EgoCross Challenge,video-qa; egocentric-video-reasoning; domain-transfer; grounded-reasoning,egocentric-video; qa
CASTLE KG Retrieval,model,2026-06,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2606.01933,https://arxiv.org/abs/2606.01933,,,Third-place CASTLE 2026 challenge framework using video knowledge graphs and adaptive hierarchical retrieval over 600+ hours of synchronized 15-camera ego/exo streams,long-context; multi-view; video-qa; temporal-reasoning,egocentric-video; exocentric-video; multimodal-annotations
Reflective Dialogue EgoCross,model,2026-05,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2605.27885,https://arxiv.org/abs/2605.27885,,,"Inference-time Teacher/Solver reflective dialogue method for EgoCross, achieving third place in the Open-source Track without fine-tuning",video-qa; egocentric-video-reasoning; domain-transfer; grounded-reasoning,egocentric-video; qa; dialogue
EARL,model,2026-05,ICML 2026,open,egocentric,2026,https://arxiv.org/abs/2605.14742,https://github.com/yuggiehk/EARL,https://github.com/yuggiehk/EARL,,"Analysis-guided RL framework for egocentric interaction reasoning and pixel grounding, with structured coarse-to-fine interaction parsing and GRPO optimization",interaction; visual-grounding; segmentation; embodied-reasoning,egocentric-video; object-masks; qa
EgoExo-WM,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.15477,https://vision.cs.utexas.edu/projects/EgoExo-WM/,,,Transforms exocentric video into egocentric video using body-pose priors to train whole-body action-conditioned egocentric world models,world-modeling; cross-view; action-conditioned-generation; human-motion,egocentric-video; exocentric-video; human-pose
MotionGRPO,model,2026-05,ICML 2026,watch,egocentric,2026,https://arxiv.org/abs/2605.05680,https://arxiv.org/abs/2605.05680,,,GRPO-based reinforcement post-training for diffusion egocentric full-body 3D motion recovery from head-mounted device signals,egocentric-3d-pose; motion-estimation; policy-optimization,egocentric-video; body-pose; human-motion
Hand Trajectory Fusion for Ego NLQ,model,2026-06,CVPR 2026 EgoVis,watch,egocentric,2026,https://arxiv.org/abs/2606.02962,https://arxiv.org/abs/2606.02962,,,"Hand-trajectory encoder and cross-attention fusion for Ego4D Natural Language Query grounding, with gains on hand-object-interaction and quantity/state queries",natural-language-query; temporal-grounding; hand-object; ego4d,egocentric-video; hand-pose; text
ACE-Ego-0,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.17200,https://arxiv.org/abs/2606.17200,,,Unifies 4.53K hours of robot/sim data with 1.48K hours of pseudo-action-labeled egocentric human video via reliability-aware weighting for VLA pretraining; evaluated on RoboCasa GR1 and RoboTwin 2.0,vla; robot-learning; imitation-learning; manipulation,egocentric-video; language; robot-actions
EgoXtreme,dataset,2026-03,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2603.25135,https://arxiv.org/abs/2603.25135,,,"Smart-glasses egocentric 6D object-pose dataset spanning industrial, sports, and rescue scenes with extreme motion blur, dynamic lighting, and heavy occlusion where current pose estimators fail",3d-4d-understanding; pose-and-body,egocentric-video; rgb
EgoForge,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.20169,https://arxiv.org/abs/2603.20169,,,Goal-directed egocentric world simulator that rolls out coherent first-person video from a single egocentric image and a high-level instruction,world-modeling; video-generation; hand-object-interaction,egocentric-video; text
EgoHOI World Model,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.13615,https://arxiv.org/abs/2603.13615,,,"Physics-informed egocentric world model that synthesizes photorealistic, contact-consistent hand-object interactions from action signals alone, without future-state inputs",world-modeling; hand-object-interaction,egocentric-video
EgoControl,model,2025-11,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2511.18173,https://arxiv.org/abs/2511.18173,,,Pose-controllable egocentric video diffusion model that generates temporally coherent first-person frames conditioned on sequences of 3D full-body poses,video-generation; world-modeling; pose-and-body,egocentric-video; body-pose
EgoFlow,model,2026-04,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2604.01421,https://arxiv.org/abs/2604.01421,,,Gradient-guided flow-matching framework that generates physically plausible 6DoF object trajectories from egocentric video using differentiable collision and smoothness constraints,3d-4d-understanding; hand-object-interaction,egocentric-video
WholeBodyVLA,model,2025-12,ICLR 2026,open,egocentric,2025,https://arxiv.org/abs/2512.11047,https://github.com/OpenDriveLab/WholebodyVLA,https://github.com/OpenDriveLab/WholebodyVLA,,Unified latent VLA that learns latent actions from action-free egocentric videos via a Latent Action Model for whole-body humanoid loco-manipulation; +21.3% over GR00T on AgiBot X2,vla; robot-learning; manipulation,egocentric-video; language; robot-actions
In-N-On,model,2025-11,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2511.15704,https://arxiv.org/abs/2511.15704,,,"Scales egocentric manipulation with 1,000+ hours of in-the-wild human egocentric video (PHSD) plus 20+ hours on-task data to train the Human0 language-conditioned flow-matching policy",robot-learning; manipulation; imitation-learning,egocentric-video; language
EgoZero,model,2025-05,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2505.20290,https://arxiv.org/abs/2505.20290,,,Trains robot manipulation policies from Project Aria smart-glasses human demonstrations with zero robot training data; ~70% zero-shot success across 7 tasks on a Franka arm,robot-learning; manipulation; imitation-learning,egocentric-video
Ego-PM,model,2025-08,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2508.19852,https://arxiv.org/abs/2508.19852,,,"Egocentric predictive world model that jointly forecasts future actions and video frames conditioned on hand trajectories via a latent diffusion model; tested on Ego4D, BridgeData, and RLBench",world-modeling; video-generation; robot-learning,egocentric-video; hand-pose; text
PEDESTRIAN,dataset,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.19190,https://arxiv.org/abs/2512.19190,,,"340 first-person pavement videos covering 29 urban sidewalk obstacle types, with deep-learning baselines for pedestrian-safety obstacle detection",detection-segmentation; ar-sensing-navigation,egocentric-video; rgb
DeltaDorsal,model,2026-01,CHI 2026,open,egocentric,2026,https://arxiv.org/abs/2601.15516,https://github.com/hilab-open-source/deltadorsal,https://github.com/hilab-open-source/deltadorsal,,Dual-stream delta encoder that uses dorsal hand-skin deformation features to improve egocentric hand pose under self-occlusion (~18% lower mean per-joint angle error),pose-and-body; hand-object-interaction,egocentric-video; hand-pose
Edge Episodic Memory QA,model,2026-02,VISAPP 2026,watch,egocentric,2026,https://arxiv.org/abs/2602.22455,https://arxiv.org/abs/2602.22455,,,On-device dual-thread MLLM pipeline (continuous video-to-text summarization plus asynchronous querying) for real-time egocentric episodic-memory QA on QAEgo4D-Closed,question-answering; memory-and-long-context,egocentric-video; text
LOME,model,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.27449,https://arxiv.org/abs/2603.27449,,,"Action-conditioned egocentric world model that generates photorealistic human-object interactions from an image, text, and per-frame body/hand actions, with realistic physical consequences",world-modeling; video-generation; hand-object-interaction,egocentric-video; body-pose; hand-pose; text
Ego-Nav Co-training,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.01951,https://arxiv.org/abs/2606.01951,,,Converts egocentric human walking videos into robot-action datasets and co-trains a VLA with robot demonstrations for mobile navigation (fruit-search task),vla; robot-learning; ar-sensing-navigation,egocentric-video; robot-actions; language
EXPLORE-Bench,benchmark,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.09731,https://arxiv.org/abs/2603.09731,,,"Egocentric scene-prediction benchmark over real first-person videos: predict the final scene state after long atomic-action sequences, with object, attribute, and relation annotations; large human-model gap",reasoning-intent-planning; question-answering; world-modeling,egocentric-video; text
Hand2World,model,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.09600,https://arxiv.org/abs/2602.09600,,,"Autoregressive egocentric world model that generates photorealistic first-person interaction video from free-space hand gestures using occlusion-invariant hand conditioning and Plucker-ray camera embeddings; best on ARCTIC, HOT3D, HOI4D",world-modeling; video-generation; hand-object-interaction,egocentric-video; hand-pose
EgoSim,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.01001,https://arxiv.org/abs/2604.01001,,,Closed-loop egocentric world simulator that generates spatially consistent first-person interaction videos with explicit 3D grounding and dynamic world-state updates for multi-stage simulation and cross-embodiment transfer,world-modeling; video-generation; robot-learning,egocentric-video; camera-pose
EgoPush,model,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.18071,https://arxiv.org/abs/2602.18071,,,"End-to-end egocentric multi-object non-prehensile rearrangement for mobile robots from a single first-person camera, using object-centric relative-pose latents and teacher-student RL with zero-shot sim-to-real transfer",robot-learning; manipulation; ar-sensing-navigation,egocentric-video
MEgoHand,model,2025-05,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2505.16602,https://arxiv.org/abs/2505.16602,,,"Multimodal egocentric hand-object interaction motion generator pairing a VLM cerebrum with a flow-matching policy over a unified 3.35M RGB-D frame, 24K interaction, 1.2K object dataset",hand-object-interaction; video-generation; pose-and-body,egocentric-video; rgb-d; text; hand-pose
EgoH4,model,2025-04,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2504.08654,https://arxiv.org/abs/2504.08654,,,"Diffusion transformer that forecasts both-hand 3D trajectories and poses from egocentric video and camera pose, including out-of-frame hands; trained on Ego-Exo4D (156K train / 34K eval sequences)",pose-and-body; hand-object-interaction; world-modeling,egocentric-video; camera-pose; hand-pose
Diff-IP2D,model,2024-05,IROS 2025,open,egocentric,2024,https://arxiv.org/abs/2405.04370,https://github.com/IRMVLab/Diff-IP2D,https://github.com/IRMVLab/Diff-IP2D,,Non-autoregressive diffusion model that jointly forecasts 2D hand trajectories and object affordances on egocentric video with explicit camera-egomotion conditioning,hand-object-interaction; pose-and-body,egocentric-video; hand-pose
IndEgo,dataset,2025-11,NeurIPS 2025,open,egocentric,2025,https://arxiv.org/abs/2511.19684,https://huggingface.co/datasets/FraunhoferIPK/IndEgo,,cc-by-4.0,"~197 hours of egocentric (plus ~97 hours exocentric) industrial collaborative-work video over assembly, logistics, inspection, and repair, with gaze, narration, sound, motion, hand pose, and point clouds",action-and-procedure; assistance-and-agents; detection-segmentation,egocentric-video; gaze; audio; motion; hand-pose; point-cloud
LaMAria,benchmark,2025-09,ICCV 2025,open,egocentric,2025,https://arxiv.org/abs/2509.26639,https://lamaria.ethz.ch,,,"City-scale egocentric visual-inertial SLAM benchmark from glasses-like wearables, hours/kilometers of trajectories with centimeter-accurate surveyed ground truth and difficulty-tiered tracks (incl. night and vehicle motion)",three-d-and-scene; ar-sensing-navigation; tracking,egocentric-video; imu; camera-pose
EgoLifter,model,2024-03,ECCV 2024,open,egocentric,2024,https://arxiv.org/abs/2403.18118,https://github.com/facebookresearch/egolifter,https://github.com/facebookresearch/egolifter,,"Open-world 3D segmentation that decomposes natural egocentric video into individual 3D objects using 3D Gaussians and SAM weak supervision, with a transient module for dynamic objects (Aria Digital Twin benchmark)",detection-segmentation; three-d-and-scene,egocentric-video; camera-pose
EgoSplat,model,2025-03,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2503.11345,https://arxiv.org/abs/2503.11345,,,"Open-vocabulary egocentric scene understanding with language-embedded 3D Gaussian splatting, using SAM2-tracked multi-view instance features and an instance-aware spatio-temporal module (gains on Aria Digital Twin)",detection-segmentation; three-d-and-scene,egocentric-video; language
3D-Aware Ego Instance Tracking,model,2024-08,ACCV 2024,watch,egocentric,2024,https://arxiv.org/abs/2408.09860,https://arxiv.org/abs/2408.09860,,,3D-aware instance segmentation and multi-object tracking in egocentric video that lifts 2D masks with scene geometry to survive rapid motion and occlusion (large ID-switch reductions on EPIC Fields),detection-segmentation; tracking; three-d-and-scene,egocentric-video; camera-pose
EgoCogNav,model,2025-11,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2511.17581,https://arxiv.org/abs/2511.17581,,,"Cognition-aware egocentric navigation framework that forecasts human trajectory and head motion from perceived path uncertainty, modeling scanning, hesitation, and backtracking, with the 6-hour CEN dataset",ar-sensing-navigation; reasoning-intent-planning; pose-and-body,egocentric-video; motion
Neck-Mounted Gaze (GLC),model,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.11669,https://arxiv.org/abs/2602.11669,,,"Transformer gaze estimator (GLC) for a shoulder-level neck-mounted camera with an out-of-bound gaze classification task and head/neck multi-view co-learning, plus a ~4-hour 8-subject dataset",ar-sensing-navigation; pose-and-body,egocentric-video; gaze
CMU-MMAC,dataset,2009-06,CMU tech report 2009,open,egocentric,2009,http://kitchen.cs.cmu.edu/,http://kitchen.cs.cmu.edu/,,not specified,"Among the earliest egocentric datasets: a multimodal kitchen-activity database with head-mounted egocentric video plus body IMUs, motion capture, and audio for 43 subjects cooking 5 recipes",action-and-procedure; detection-segmentation,egocentric-video; imu; motion; audio
First-Person Social Interactions,dataset,2012-06,CVPR 2012,open,egocentric,2012,http://ai.stanford.edu/~alireza/publication/CVPR12.pdf,http://ai.stanford.edu/~alireza/publication/CVPR12.pdf,,not specified,"The first egocentric social-interaction dataset: day-long head-mounted video of 8 subjects at a theme park, annotated for social interactions, roles, attention, and turn-taking",audio-and-social; reasoning-intent-planning; detection-segmentation,egocentric-video
BEOID,dataset,2014-09,BMVC 2014,open,egocentric,2014,https://dimadamen.github.io/BEOID/,https://dimadamen.github.io/BEOID/,,not specified,"Gaze-tracked egocentric video of 8 users interacting with objects across 6 everyday locations (kitchen, workspace, printer, corridor, gym) for discovering task-relevant objects and interaction modes",hand-object-interaction; action-and-procedure; detection-segmentation,egocentric-video; gaze
DreamDojo,model,2026-02,ICML 2026,open,egocentric,2026,https://arxiv.org/abs/2602.06949,https://github.com/NVIDIA/DreamDojo,https://github.com/NVIDIA/DreamDojo,,"Generalist robot world model pretrained on 44K hours of egocentric human video using continuous latent actions; distilled to real-time (10.8 FPS) for teleoperation, policy evaluation, and model-based planning",world-modeling; robot-learning; video-generation,egocentric-video; latent-actions
UniDex,model,2026-03,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2603.22264,https://arxiv.org/abs/2603.22264,,,"Robot foundation suite for universal dexterous hand control learned from egocentric human videos, transferring human manipulation priors to multi-fingered robot hands",robot-learning; dexterous-manipulation; vla,egocentric-video; hand-pose
DexGloveHOI,dataset,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.21714,https://arxiv.org/abs/2605.21714,,,100K+ synchronized egocentric vision and on-glove IMU samples with marker-based mocap ground-truth 3D hand pose across dexterous daily object manipulation (introduced with the AVI-HT vision-IMU fusion method),pose-and-body; hand-object-interaction; 3d-4d-understanding,egocentric-video; imu; hand-pose
DexWM,model,2025-12,CVPR 2026,watch,egocentric,2025,https://arxiv.org/abs/2512.13644,https://arxiv.org/abs/2512.13644,,,"Dexterous interaction world model that learns fine-grained hand-object manipulation from finger keypoints in 900+ hours of egocentric human and robot video, with zero-shot transfer to a Franka-Allegro hand",world-modeling; hand-object-interaction; robot-learning,egocentric-video; hand-pose
H2R,model,2025-05,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2505.11920,https://arxiv.org/abs/2505.11920,,,"Human-to-robot data augmentation that turns egocentric human videos into robot-centric training data by retargeting hand pose, inpainting human limbs, and compositing rendered robot arms for manipulation pre-training",robot-learning; imitation-learning; manipulation,egocentric-video; hand-pose
HumanNet,dataset,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.06747,https://arxiv.org/abs/2605.06747,,,"One-million-hour human-centric video corpus spanning first-person and third-person perspectives, with interaction-centric captions, motion descriptions, hand/body signals, and a reported 1,000-hour egocentric subset used for VLA-style validation",foundation-video; representation-learning; embodied-ai; action-and-procedure; hand-object-interaction; robot-learning; vla,egocentric-video; exocentric-video; captions; motion-descriptions; hand-pose; body-pose
HumanScale,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.20521,https://arxiv.org/abs/2606.20521,,,"Pretraining study and recipe showing that filtered, labeled egocentric human video can outperform teleoperated real-robot data for embodied pretraining (lower action-prediction loss; large in- and out-of-distribution success gains)",robot-learning; vla; representation-learning,egocentric-video
Do as I Do,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.19333,https://arxiv.org/abs/2606.19333,,,Reconstructs and retargets human hand-object interactions from in-the-wild monocular RGB video into executable dexterous robot manipulation trajectories,robot-learning; dexterous-manipulation; hand-object-interaction,egocentric-video; hand-pose
EgoCS-400K,dataset,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.18180,https://arxiv.org/abs/2606.18180,,,"400K+ first-person Counter-Strike gameplay videos (10K hours, 1K+ matches, 40K rounds, 13 maps) with aligned actions, player state, camera motion, view direction, and game events for interactive world-model training",world-modeling; video-generation; action-and-procedure,egocentric-video; action-labels; camera-pose
UNIEGO,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.20559,https://arxiv.org/abs/2606.20559,,,"Unified egocentric encoder trained through hierarchical multi-teacher distillation from nine teachers across ego-exo viewpoints, RGB, depth, skeleton, and foundation-model representations",representation-learning; action-recognition; video-retrieval; action-segmentation,egocentric-video; exocentric-video; depth; skeleton
HT-Bench / HandTouch,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.19161,https://arxiv.org/abs/2606.19161,,,"Dexterous full-hand tactile benchmark pairing egocentric RGB with 10M RGB frames and 7.8M tactile frames across 226 tasks, plus the HandTouch vision-tactile encoder",contact-understanding; hand-object-interaction; robot-learning; evaluation,egocentric-video; tactile; hand-pose
EPIC-Contact / HOPformer,dataset,2026-06,ECCV 2026,open,egocentric,2026,https://arxiv.org/abs/2606.30598,https://sid2697.github.io/epic-contact/,https://github.com/Sid2697/HOPformer,cc-by-nc-4.0,"EPIC-KITCHENS-derived in-the-wild egocentric 3D hand-object pose dataset with 2.3K clips, 62.3K frames, nine object categories, dense bijective 3D hand-object contact correspondences, posed meshes, plus HOPformer code and checkpoints",hand-object-pose; 3d-hand-pose; contact-understanding; hand-object-interaction; 3d-hand-object-tracking,egocentric-video; hand-pose; object-pose; hand-meshes; object-meshes; contact
GhostHandEgoNet,model,2026-06,IJPRAI 2026,watch,egocentric,2026,https://doi.org/10.1142/s0218001426550116,https://doi.org/10.1142/s0218001426550116,,,"Lightweight egocentric hand-pose and hand-hygiene monitoring pipeline using Ghost convolutions with attention, reporting 2.56M parameters and robust embedded deployment under occlusion",3d-hand-pose; hand-object-interaction; medical-assistance; efficient-inference,egocentric-video; hand-pose
Towards Egocentric 3D Hand Pose Estimation in Unseen Domains,model,2026-03,WACV 2026,watch,egocentric,2026,https://doi.org/10.1109/wacv61042.2026.00560,https://doi.org/10.1109/wacv61042.2026.00560,,,Egocentric 3D hand-pose estimation study focused on unseen-domain generalization for first-person hand tracking,3d-hand-pose; hand-tracking; domain-generalization; ar-vr,egocentric-video; hand-pose
Leveraging Synthetic Data for Egocentric HOI Detection,model,2026-05,IJCV 2026,watch,egocentric,2026,https://doi.org/10.1007/s11263-026-02838-8,https://doi.org/10.1007/s11263-026-02838-8,,,"Studies synthetic-data augmentation for egocentric hand-object interaction detection on VISOR, EgoHOS, and ENIGMA-51, especially when real labels are scarce",hoi-detection; hand-object-interaction; synthetic-data; data-augmentation; domain-generalization,egocentric-video; synthetic-data; hand-masks; object-masks
Person Identification from Egocentric HOI,model,2026-02,CCF TPCI 2026,watch,egocentric,2026,https://doi.org/10.1007/s42486-025-00215-x,https://doi.org/10.1007/s42486-025-00215-x,,,Person-identification study using 3D hand pose from egocentric human-object interactions as an identity signal,hand-object-interaction; 3d-hand-pose; person-identification; behavior-understanding,egocentric-video; hand-pose; human-demonstrations
Industrial Egocentric HOI Detection,model,2026-01,ICIAP 2025 Workshops,watch,egocentric,2026,https://doi.org/10.1007/978-3-032-11317-7_38,https://doi.org/10.1007/978-3-032-11317-7_38,,,Real-time egocentric hand-object interaction detection system for industrial-domain workflows,hoi-detection; hand-object-interaction; industrial-assembly; efficient-inference,egocentric-video; hand-masks; object-masks
GlovEgo-HOI,model,2026,VISAPP 2026,watch,egocentric,2026,https://doi.org/10.5220/0014466400004084,https://doi.org/10.5220/0014466400004084,,,Industrial egocentric HOI detection work bridging the synthetic-to-real gap for gloved hand-object interaction scenarios,hoi-detection; hand-object-interaction; synthetic-data; domain-adaptation; industrial-assembly,egocentric-video; synthetic-data; hand-masks; object-masks
Causal-Inspired Fourier Egocentric Action Recognition,model,2026-05,IEEE TCSVT 2026,watch,egocentric,2026,https://doi.org/10.1109/tcsvt.2026.3653125,https://doi.org/10.1109/tcsvt.2026.3653125,,,Causal-inspired Fourier representation learning for wearable IMU signals and egocentric action recognition,action-recognition; representation-learning; wearable-sensing; sensor-fusion,imu; wearable-sensing; egocentric-video
Screen Detection from Egocentric Streams,model,2026,IEEE TMM 2026,watch,egocentric,2026,https://doi.org/10.1109/tmm.2026.3660180,https://doi.org/10.1109/tmm.2026.3660180,,,Screen detection from egocentric image streams using a multi-view vision-language model,object-detection; scene-text; assistive-systems; wearable-sensing,egocentric-images; wearable-sensing; text
SmartSeg,model,2026,Pervasive and Mobile Computing 2026,watch,egocentric,2026,https://doi.org/10.1016/j.pmcj.2026.102223,https://doi.org/10.1016/j.pmcj.2026.102223,,,Non-parametric temporal segmentation method for wearable-camera video streams,temporal-segmentation; procedure-understanding; lifelogging; wearable-sensing,egocentric-video; wearable-sensing
Hand-4DGS,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.19156,https://jeongminb.github.io/hand-4dgs/,,,"Feed-forward 3D Gaussian Splatting framework for dynamic 4D hand reconstruction from egocentric videos, targeting fast inference around 60 FPS on H2O and ARCTIC-style capture",3d-4d-understanding; pose-and-body; ar-vr,egocentric-video; hand-pose; hand-meshes
Motion-Focused Latent Action VLA,model,2026-06,IROS 2026,watch,egocentric,2026,https://arxiv.org/abs/2606.18955,https://arxiv.org/abs/2606.18955,,,Cross-embodiment VLA training recipe that extracts motion-focused latent actions from unlabeled human EgoVideos with a hybrid disentangled VQ-VAE and adapts with 50 downstream trajectories,vla; robot-learning; cross-embodiment-transfer; imitation-learning,egocentric-video; language; robot-actions
VEGA / VEGA-Bench,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.18426,https://arxiv.org/abs/2606.18426,,,"Navigation VLA method and benchmark using in-the-wild egocentric videos, monocular geometry, 250K scenes, and about 5M navigation goals for obstacle-aware trajectory supervision",vla; ar-sensing-navigation; robot-learning; evaluation,egocentric-video; scene-geometry; trajectories; language
Wh0,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.22136,https://chenyt31.github.io/wh0.github.io/,,,Generates 50K episodes of synthetic egocentric human-hand manipulation video (WM-H) and converts them into robot-trainable dexterous action supervision using hand-motion reconstruction,vla; dexterous-manipulation; world-modeling; robot-learning,egocentric-video; text; robot-actions
Divide Deliberate Decide,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.17627,https://arxiv.org/abs/2606.17627,,,"Fully local zero-shot multi-agent VLM framework for fine-grained egocentric action recognition, combining orchestrated video chunks, heterogeneous VLM specialists, peer consultation, and Borda aggregation",action-recognition; egocentric-video-reasoning; grounded-reasoning,egocentric-video; text
VL-MemKnG / WalkieKnowledgeT+,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.17183,https://arxiv.org/abs/2606.17183,,,"Hybrid spatio-temporal knowledge-graph and segment-memory framework for QA over long egocentric navigation trajectories, with WalkieKnowledgeT+ for temporally distributed reasoning",ar-sensing-navigation; memory-and-long-context; question-answering; spatial-reasoning,egocentric-video; qa; trajectory; scene-graphs
EgoPhys,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.16202,https://hjhyunjinkim.github.io/EgoPhys,,,Learns deformable-object physics digital twins from egocentric RGB-only interaction video using inverse-physics codebooks and an egocentric deformable-object interaction dataset,world-modeling; robot-learning; manipulation; 3d-4d-understanding,egocentric-video; rgb; physics; robot-actions
EPIC Efficient Egocentric Perception,toolkit,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.15859,https://arxiv.org/abs/2606.15859,,,"Algorithm-hardware co-optimization framework for smart AR glasses that uses gaze, pose, and inertial signals to retain informative high-resolution egocentric perception while reducing memory and energy cost",ar-vr; wearable-sensing; embodied-ai; efficient-inference,egocentric-video; gaze; pose; imu
Temporal Action Graphs for Ego VLMs,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.15417,https://arxiv.org/abs/2606.15417,,,"In-context egocentric action-recognition pipeline that converts video into dense narratives and temporal action graphs, improving few-shot reasoning on EGTEA and EPIC-KITCHENS-100 with open-weight VLMs",action-recognition; symbolic-reasoning; video-language,egocentric-video; text; temporal-graphs
V-RAGBench / CARVE,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.13141,https://arxiv.org/abs/2606.13141,,,"Long egocentric-video RAG benchmark and chunk-adaptive retrieval method using query, evidence-chunk, answer triplets to evaluate retrieval and generation separately",memory-and-long-context; retrieval; question-answering; evaluation,egocentric-video; qa; text
ReRe Cross-View Revisiting,model,2026-06,ICML 2026,watch,egocentric,2026,https://arxiv.org/abs/2606.11683,https://zhenjiemao.github.io/ReRe/,,,Training-free egocentric spatial-reasoning framework that revisits first-person video conclusions through synthesized complementary novel views on VSI-Bench and STI-Bench,spatial-reasoning; 3d-4d-understanding; video-qa,egocentric-video; novel-view-video; 3d
EDITH,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.10276,https://project-edith.github.io,https://project-edith.github.io,,"Hierarchical robot-policy framework that streams first-person view, gaze, and speech from smart glasses to infer intent, produce grounded subtasks, and reduce communication burden in human-robot interaction",robot-learning; hri; gaze; language-grounding,egocentric-video; gaze; speech; robot-actions
Eva-3M / EvaPose,dataset,2026-02,CVPR 2026,watch,egocentric,2026,https://arxiv.org/abs/2602.23618,https://arxiv.org/abs/2602.23618,,,"Large-scale egocentric visibility-aware human-pose dataset with over 3.0M frames and 435K keypoint-visibility annotations, plus the EvaPose visibility-aware HPE method",egocentric-3d-pose; pose-and-body; ar-vr,egocentric-video; body-pose; keypoint-visibility
SAW-Bench,benchmark,2026-02,ICML 2026,open,egocentric,2026,https://arxiv.org/abs/2602.16682,https://arxiv.org/abs/2602.16682,https://github.com/UCSB-AI/SAW-Bench,cc-by-nc-4.0,"ICML 2026 situated-awareness benchmark with 786 Ray-Ban Meta smart-glasses videos and 2,071 human-annotated QA pairs across localization, direction, reverse planning, shape, affordance, and memory",spatial-reasoning; situated-awareness; video-qa; evaluation,egocentric-video; qa
ARGaze,model,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.05132,https://arxiv.org/abs/2602.05132,,,"Autoregressive transformer for online egocentric gaze estimation, using current visual features and bounded recent gaze context for causal streaming inference",gaze; ar-vr; wearable-sensing,egocentric-video; gaze
STAformer++ Affordance-Aware Anticipation,model,2026-02,IEEE TPAMI,watch,egocentric,2026,https://arxiv.org/abs/2602.14837,https://arxiv.org/abs/2602.14837,,,Short-term object-interaction anticipation method integrating image-video attention with environment affordance memory and interaction hotspots on Ego4D and curated EPIC-KITCHENS STA labels,anticipation; hand-object-interaction; affordance; action-and-procedure,egocentric-video; object-boxes; hand-trajectories
WristPP,model,2026-02,CHI 2026 submission,watch,egocentric,2026,https://arxiv.org/abs/2603.00606,https://zhenqis123.github.io/WristPP/,,,"Wrist-worn wide-FOV RGB system for real-time 3D hand pose and per-vertex pressure estimation, evaluated on a 133K-frame, 20-subject pose-pressure dataset",pose-and-body; contact-understanding; ar-vr,wrist-camera-video; hand-pose; pressure
Egocentric Co-Pilot,toolkit,2026-03,WWW 2026,watch,egocentric,2026,https://arxiv.org/abs/2603.01104,https://arxiv.org/abs/2603.01104,,,"Web-native neuro-symbolic smart-glasses agent framework with Temporal Chain-of-Thought, hierarchical context compression, speech/gaze intent parsing, and WebRTC/WebSocket streaming for assistive egocentric AI",tool-using-agents; streaming-assistant; memory-and-long-context; assistive-videoqa,egocentric-video; speech; gaze; web-tools
Gazeify Then Voiceify,model,2026-01,IUI 2026,watch,egocentric,2026,https://arxiv.org/abs/2601.19281,https://arxiv.org/abs/2601.19281,,,"Displayless smart-glasses interaction system for physical-object referencing through gaze selection, object masks, voice descriptions, and conversational correction",referential; ar-vr; interactive-assistance; visual-grounding,egocentric-video; gaze; speech; object-masks
EGAgent,model,2026-01,arXiv,open,egocentric,2026,https://arxiv.org/abs/2601.18157,https://github.com/facebookresearch/egagent,https://github.com/facebookresearch/egagent,,"Agentic very-long-video understanding framework for all-day wearable egocentric streams, using entity scene graphs, hybrid visual/audio search, and planning tools on EgoLifeQA and Video-MME Long",memory-and-long-context; question-answering; tool-using-agents; audio-visual-reasoning,egocentric-video; audio; scene-graphs; qa
Wearable Product Localization,model,2026-01,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2601.12486,https://arxiv.org/abs/2601.12486,,,"Multimodal wearable assistive system for blind and low-vision product search, navigation, and correction, combining YOLO-World, VLM descriptions, spatialized sonification, and retrieval feedback",assistive-videoqa; ar-sensing-navigation; object-detection; interactive-assistance,egocentric-video; audio; language
VisionClaw,toolkit,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.03486,https://arxiv.org/abs/2604.03486,,,Always-on wearable AI-agent system on Ray-Ban Meta smart glasses that couples live egocentric perception with speech-driven OpenClaw task execution for situated hands-free actions,tool-using-agents; streaming-assistant; interactive-assistance; wearable-sensing,egocentric-video; speech; web-tools
ActiveGlasses,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.08534,https://arxiv.org/abs/2604.08534,,,"Smart-glasses active-vision system for learning robot manipulation from ego-centric human demonstrations, using a stereo head-mounted camera for collection and robot-mounted active perception at deployment",robot-learning; imitation-learning; cross-embodiment-transfer; manipulation,egocentric-video; stereo-video; point-clouds; robot-actions
INSIGHT,model,2025-08,AAAI 2026,watch,egocentric,2025,https://arxiv.org/abs/2508.01742,https://arxiv.org/abs/2508.01742,,,"Intention-guided cognitive reasoning framework for egocentric long-term action anticipation, using hand-object cues, verb-noun dependencies, and RL-style reasoning over Ego4D, EPIC-KITCHENS-55, and EGTEA Gaze+",anticipation; action-and-procedure; hand-object-interaction; reasoning-intent-planning,egocentric-video; hand-pose; object-boxes
NS-iHOS / WISH,model,2025-09,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2509.26004,https://arxiv.org/abs/2509.26004,,,"Weakly supervised egocentric in-hand object segmentation from natural-language narrations, distilling narration cues into WISH for hand-object association and segmentation without narration at inference",hand-object-interaction; detection-segmentation; video-language,egocentric-video; narrations; object-masks
Exo2Ego / Ego-ExoClip,model,2025-03,AAAI 2026,watch,egocentric,2025,https://arxiv.org/abs/2503.09143,https://arxiv.org/abs/2503.09143,,,"Exocentric-to-egocentric MLLM transfer with Ego-ExoClip, a 1.1M synchronized ego-exo clip-text pretraining set from Ego-Exo4D, plus EgoIT instruction tuning",video-language; ego-exo; cross-view; representation-learning,egocentric-video; exocentric-video; text
EgoWorld,model,2025-06,ICLR 2026,watch,egocentric,2025,https://arxiv.org/abs/2506.17896,https://redorangeyellowy.github.io/EgoWorld/,,,"Exocentric-to-egocentric generation framework that reconstructs egocentric views from point clouds, 3D hand poses, and text, evaluated on H2O, TACO, Assembly101, and Ego-Exo4D",world-modeling; video-generation; cross-view; hand-object-interaction,egocentric-video; exocentric-video; point-clouds; hand-pose; text
Aria STDR,dataset,2025-07,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2507.16330,https://arxiv.org/abs/2507.16330,,,"Custom Project Aria smart-glasses scene-text detection and recognition dataset studying lighting, distance, resolution, and gaze-guided processing for egocentric OCR",scene-text; ar-vr; wearable-sensing; detection-segmentation,egocentric-video; gaze; scene-text
RaycastGrasp,toolkit,2025-10,ICIR 2025,watch,egocentric,2025,https://arxiv.org/abs/2510.22113,https://arxiv.org/abs/2510.22113,,,"Egocentric gaze-guided robotic manipulation interface using a wearable MR headset, augmented visual cues, pretrained vision models, and a robotic arm for object retrieval",hri; robot-learning; gaze; interactive-assistance,egocentric-video; gaze; robot-actions
Fish2Mesh,model,2025-03,ICCV 2025,watch,egocentric,2025,https://arxiv.org/abs/2503.06089,https://arxiv.org/abs/2503.06089,,,"Fisheye-aware transformer for 3D human mesh recovery from egocentric vision, using ego-specific position embeddings and weak supervision from third-person cameras",egocentric-3d-pose; pose-and-body; ar-vr,egocentric-video; body-pose; human-motion
GazeLLM,model,2025-03,Augmented Humans 2025,watch,egocentric,2025,https://arxiv.org/abs/2504.00221,https://arxiv.org/abs/2504.00221,,,Gaze-focused multimodal LLM pipeline that reduces first-person video processing by routing high-resolution analysis to human attention regions while preserving task comprehension,video-language; gaze; wearable-sensing; efficient-inference,egocentric-video; gaze; audio; text
LookOut / Aria Navigation Dataset,dataset,2025-08,ICCV 2025,watch,egocentric,2025,https://arxiv.org/abs/2508.14466,https://arxiv.org/abs/2508.14466,,,"Project Aria real-world humanoid egocentric navigation dataset and framework for predicting future 6D head-pose trajectories, with 4 hours of navigation recordings",ar-sensing-navigation; trajectory-prediction; robot-learning; pose-and-body,egocentric-video; head-pose; camera-pose
ECHO,model,2025-08,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2508.21556,https://arxiv.org/abs/2508.21556,,,"Ego-centric HOI diffusion framework recovering human pose, object motion, and contact dynamics from sparse smart-glasses and smartwatch tracking signals",hand-object-interaction; contact-understanding; pose-and-body; ar-vr,egocentric-video; head-pose; wrist-trajectory; object-motion; contact
DMC3 Ego VideoQA,model,2025-10,ACM MM 2025,watch,egocentric,2025,https://arxiv.org/abs/2510.20285,https://arxiv.org/abs/2510.20285,,,"Dual-modal counterfactual contrastive framework for egocentric VideoQA, constructing textual and visual counterfactual samples around events and hand-object interactions",question-answering; video-qa; hand-object-interaction; contrastive-learning,egocentric-video; qa; text
Uni-Hand,model,2025-11,T-PAMI 2026,open,egocentric,2025,https://arxiv.org/abs/2511.12878,https://github.com/IRMVLab/UniHand,https://github.com/IRMVLab/UniHand,,"Universal egocentric hand motion forecasting framework for multi-modal input, 2D/3D wrist and finger waypoints, head-hand motion synergy, and downstream affordance tasks",hand-object-interaction; pose-and-body; anticipation; robot-learning,egocentric-video; hand-pose; text
SFHand / EgoHaFL,model,2025-11,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2511.18127,https://arxiv.org/abs/2511.18127,,,Streaming language-guided 3D hand-forecasting framework plus EgoHaFL dataset with synchronized 3D hand poses and language instructions for embodied manipulation,hand-object-interaction; pose-and-body; robot-learning; streaming-assistant,egocentric-video; hand-pose; language-instructions
OpenTouch,dataset,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.16842,https://opentouch-tactile.github.io/,,,"In-the-wild egocentric full-hand tactile dataset with 5.1 hours of synchronized video-touch-pose data, 2,900 curated clips, text annotations, and retrieval/classification benchmarks",contact-understanding; hand-object-interaction; retrieval; robot-learning,egocentric-video; tactile; hand-pose; text
mmEgoHand,model,2025-01,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2501.13805,https://arxiv.org/abs/2501.13805,,,"Head-mounted egocentric hand pose and gesture-recognition system combining millimeter-wave radar and IMU to support VR interaction, skill digitization, and robotic teleoperation",pose-and-body; gesture-recognition; ar-vr; robot-learning,imu; radar; hand-pose
AiGet,toolkit,2025-01,CHI 2025,watch,egocentric,2025,https://arxiv.org/abs/2501.16240,https://arxiv.org/abs/2501.16240,,,"Smart-glasses AI assistant for informal learning in everyday moments, using real-time gaze patterns, environmental context, user profiles, and LLM-generated knowledge prompts",interactive-assistance; wearable-sensing; tool-using-agents,egocentric-video; gaze; language
Ego4o,model,2025-04,CVPR 2025,watch,egocentric,2025,https://arxiv.org/abs/2504.08449,https://arxiv.org/abs/2504.08449,,,"Omni-modal egocentric human motion capture and understanding framework for VR/AR headsets, smart glasses, phones, and watches with egocentric images, sparse IMUs, and motion-language descriptions",pose-and-body; motion-language; ar-vr; wearable-sensing,egocentric-video; imu; text; human-motion
EgoExo-Gen,model,2025-04,ICLR 2025,watch,egocentric,2025,https://arxiv.org/abs/2504.11732,https://arxiv.org/abs/2504.11732,,,"Cross-view egocentric video prediction from exocentric video, first ego frame, and text instructions, using predicted hand-object interaction masks and diffusion generation",world-modeling; video-generation; ego-exo; hand-object-interaction,egocentric-video; exocentric-video; text; object-masks
Reading in the Wild,dataset,2025-05,NeurIPS 2025,watch,egocentric,2025,https://arxiv.org/abs/2505.24848,https://www.projectaria.com/datasets/reading-in-the-wild/,,,"Project Aria smart-glasses dataset for reading recognition with 100 hours of reading and non-reading video, egocentric RGB, eye gaze, head pose, and reading-type labels",reading; wearable-sensing; ar-vr; activity-recognition,egocentric-video; gaze; head-pose
Ego 6DoF Object Trajectories,model,2025-06,CVPR 2025,watch,egocentric,2025,https://arxiv.org/abs/2506.03605,https://arxiv.org/abs/2506.03605,,,"Framework extracting 6DoF manipulation trajectories from ego/exo video datasets and generating object manipulation trajectories from action descriptions in egocentric vision, evaluated with HOT3D",hand-object-interaction; robot-learning; manipulation; 3d-4d-understanding,egocentric-video; exocentric-video; object-pose; text
HEADS-UP,dataset,2024-09,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2409.20324,https://arxiv.org/abs/2409.20324,,,Head-mounted egocentric dataset for pedestrian trajectory prediction and collision-risk assessment in blind-assistance systems,ar-sensing-navigation; trajectory-prediction; blind-assistance; assistive-systems,egocentric-video; trajectories
Embodied VideoAgent,model,2024-12,ICCV 2025,watch,egocentric,2024,https://arxiv.org/abs/2501.00358,https://embodied-videoagent.github.io/,,,"LLM/VLM agent that builds persistent 3D scene memory from egocentric video plus embodied depth and pose sensors for dynamic scene reasoning, planning, and robot manipulation",memory-and-long-context; spatial-reasoning; embodied-ai; robot-learning,egocentric-video; depth; pose; scene-graphs
SoundingActions,model,2024-04,CVPR 2024,watch,egocentric,2024,https://arxiv.org/abs/2404.05206,https://vision.cs.utexas.edu/projects/soundingactions,,,Self-supervised audio-language-vision embedding that learns how actions sound from narrated egocentric videos using contrastive consensus over Ego4D and EPIC-Sounds,audio-visual-reasoning; action-and-procedure; representation-learning,egocentric-video; audio; narrations
EgoChoir,model,2024-05,NeurIPS 2024,watch,egocentric,2024,https://arxiv.org/abs/2405.13659,https://yyvhang.github.io/EgoChoir/,,,"Egocentric 3D human-object interaction region model inferring 3D contact and object affordance from egocentric video, head motion, and 3D object cues",hand-object-interaction; contact-understanding; 3d-4d-understanding; affordance,egocentric-video; head-pose; object-meshes; contact
ObjectRelator,model,2024-11,ICCV 2025 Highlight,watch,egocentric,2024,https://arxiv.org/abs/2411.19083,https://arxiv.org/abs/2411.19083,,,Cross-view ego-exo object relation and correspondence model with multimodal condition fusion and self-supervised cross-view object alignment,ego-exo; cross-view; object-grounding; detection-segmentation,egocentric-video; exocentric-video; object-masks; text
HUX,toolkit,2024-07,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2407.19492,https://arxiv.org/abs/2407.19492,,,"Heads Up eXperience always-on AI companion concept for smart glasses and XR headsets, integrating eye gaze, surrounding context, verbal context, memory storage, and task assistance",interactive-assistance; memory-and-long-context; wearable-sensing; tool-using-agents,egocentric-video; gaze; speech; memory
EgoZAR,model,2024-09,Pattern Recognition Letters,watch,egocentric,2024,https://arxiv.org/abs/2409.14205,https://gabrielegoletto.github.io/EgoZAR/,,,Egocentric zone-aware action-recognition method that decouples activity-centric zone priors from scene-specific appearance to improve cross-domain transfer on EPIC-KITCHENS-100 and Argo1M,action-recognition; domain-transfer; scene-understanding,egocentric-video; scene-labels
TouchInsight,model,2024-10,UIST 2024,watch,egocentric,2024,https://arxiv.org/abs/2410.05940,https://arxiv.org/abs/2410.05940,,,"Real-time mixed-reality touch and text input pipeline that detects touch moment, finger, and touch location from egocentric hand tracking on arbitrary physical surfaces",ar-vr; contact-understanding; hand-object-interaction; interactive-assistance,egocentric-video; hand-pose; contact
EffHandEgoNet,model,2024-04,FG 2024,watch,egocentric,2024,https://arxiv.org/abs/2404.09308,https://arxiv.org/abs/2404.09308,,,"Egocentric 2D hand-pose and action-recognition method for smart-glasses RGB input, including EffHandNet and EffHandEgoNet evaluated on H2O and FPHA",hand-object-interaction; pose-and-body; action-recognition,egocentric-video; hand-pose; object-boxes
Untrimmed Action Anticipation,model,2022-02,ICIAP 2022,watch,egocentric,2022,https://arxiv.org/abs/2202.04132,https://arxiv.org/abs/2202.04132,,,Reframes egocentric action anticipation for untrimmed streams rather than short pre-cut observation windows,anticipation; action-and-procedure; temporal-segmentation,egocentric-video
EgoGAN,model,2022-03,ECCV 2022,watch,egocentric,2022,https://arxiv.org/abs/2203.11305,https://arxiv.org/abs/2203.11305,,,Generative model for forecasting future hand masks from egocentric video under stochastic head and hand motion,hand-segmentation; anticipation; hand-object-interaction,egocentric-video; hand-masks
HOI Forecast / Interaction Hotspots,model,2022-04,CVPR 2022,open,egocentric,2022,https://arxiv.org/abs/2204.01696,https://stevenlsw.github.io/hoi-forecast,,,Forecasts future hand trajectories and interaction hotspots on next-active objects from egocentric video,hand-forecasting; anticipation; affordance; hand-object-interaction,egocentric-video; hand-trajectories; object-masks
SOS Handled Objects,model,2022-04,ECCV 2022,watch,egocentric,2022,https://arxiv.org/abs/2204.04796,https://arxiv.org/abs/2204.04796,,,Self-supervised learning over sets of handled objects to improve egocentric action recognition,action-recognition; hand-object-interaction; representation-learning,egocentric-video; object-boxes
I-CVAE LTA,model,2022-07,WACV 2023,open,egocentric,2022,https://arxiv.org/abs/2207.12080,https://evm7.github.io/icvae-page/,,,Intention-conditioned variational model for long-term human egocentric action forecasting; reported Ego4D LTA challenge wins,anticipation; intent; long-horizon-memory; action-and-procedure,egocentric-video; narrations
UmeTrack,model,2022-10,SIGGRAPH Asia 2022,watch,egocentric,2022,https://arxiv.org/abs/2211.00099,https://arxiv.org/abs/2211.00099,,,Unified multi-view end-to-end 3D hand tracking for VR headsets and hand-interaction systems,3d-hand-pose; wearable-interaction; ar-vr,egocentric-video; hand-pose
InternVideo-Ego4D,model,2022-11,Ego4D Workshop 2022,watch,egocentric,2022,https://arxiv.org/abs/2211.09529,https://arxiv.org/abs/2211.09529,,,"Pack of champion Ego4D challenge solutions spanning episodic memory, forecasting, hand-object, and audio/social tracks",benchmark; evaluation; action-recognition; temporal-grounding,egocentric-video; audio; narrations
EgoT2 / Egocentric Video Task Translation,model,2022-12,CVPR 2023 Highlight,open,egocentric,2022,https://arxiv.org/abs/2212.06301,https://vision.cs.utexas.edu/projects/egot2/,,,Task-translation framework that uses one egocentric video task to supervise or transfer to another in wearable-video settings,video-language; task-suite; representation-learning; action-and-procedure,egocentric-video; text
EgoLoc,model,2022-12,ICCV 2023,watch,egocentric,2022,https://arxiv.org/abs/2212.06969,https://arxiv.org/abs/2212.06969,,,Revisits Ego4D visual-query 3D object localization with stronger camera-pose estimation and localization baselines,3d; object-grounding; visual-grounding; tracking,egocentric-video; camera-pose; 3d-point-clouds
NaQ,model,2023-01,CVPR 2023,watch,egocentric,2023,https://arxiv.org/abs/2301.00746,https://arxiv.org/abs/2301.00746,,,Uses narrations as implicit queries to supervise episodic-memory retrieval in long egocentric video,episodic-memory; natural-language-query; retrieval; video-language,egocentric-video; narrations; text
Ego-Only,model,2023-01,ICCV 2023,watch,egocentric,2023,https://arxiv.org/abs/2301.01380,https://arxiv.org/abs/2301.01380,,,Egocentric action-detection approach that removes reliance on exocentric pretraining transfer,action-detection; action-recognition; representation-learning,egocentric-video
EgoDistill,model,2023-01,tech report,open,egocentric,2023,https://arxiv.org/abs/2301.02217,https://vision.cs.utexas.edu/projects/egodistill/,,,Distills expensive egocentric clip features into efficient models by reconstructing heavy representations from head-motion signals,efficient-inference; representation-learning; wearable-sensing,egocentric-video; head-pose
TransFusion,model,2023-01,IEEE TCSVT 2024,watch,egocentric,2023,https://arxiv.org/abs/2301.09209,https://arxiv.org/abs/2301.09209,,,Multimodal object-interaction anticipation model that summarizes past egocentric context with natural-language descriptions,anticipation; next-active-object; video-language; hand-object-interaction,egocentric-video; text; object-boxes
Fine-Grained Affordance Annotation,benchmark,2023-02,WACV 2023,watch,egocentric,2023,https://arxiv.org/abs/2302.03292,https://arxiv.org/abs/2302.03292,,,"Fine-grained affordance annotation protocol for egocentric hand-object interaction videos, extending earlier EPIC/Ego4D workshop annotation work",affordance; hand-object-interaction; detection-segmentation,egocentric-video; object-masks; hand-masks
Next Active Object Anticipation,model,2023-02,IEEE Access 2024,watch,egocentric,2023,https://arxiv.org/abs/2302.06358,https://arxiv.org/abs/2302.06358,,,Anticipates next-active-object locations before contact in egocentric video,next-active-object; anticipation; hand-object-interaction,egocentric-video; object-boxes
MINOTAUR,model,2023-02,arXiv,watch,egocentric,2023,https://arxiv.org/abs/2302.08063,https://arxiv.org/abs/2302.08063,,,"Unified video grounding model for multimodal queries over tasks such as action detection, visual queries, and spatio-temporal language grounding",temporal-grounding; visual-grounding; object-grounding; video-language,egocentric-video; text; object-boxes
EgoViT,model,2023-03,arXiv,watch,egocentric,2023,https://arxiv.org/abs/2303.08920,https://arxiv.org/abs/2303.08920,,,Pyramid video transformer with dynamic class-token generation for egocentric action recognition,action-recognition; hand-object-interaction; representation-learning,egocentric-video
AV-EgoGaze Anticipation,model,2023-05,ECCV 2024,watch,egocentric,2023,https://arxiv.org/abs/2305.03907,https://arxiv.org/abs/2305.03907,,,Audio-visual egocentric gaze-anticipation model using video and audio cues to forecast future attention,gaze; anticipation; audio-visual; ar-vr,egocentric-video; audio; gaze
AntGPT,model,2023-07,ICLR 2024,watch,egocentric,2023,https://arxiv.org/abs/2307.16368,https://arxiv.org/abs/2307.16368,,,Uses large language models to reason over common future event sequences and improve long-term action anticipation,anticipation; long-horizon-memory; planning; action-and-procedure,egocentric-video; text
Encode-Store-Retrieve,model,2023-08,ISMAR 2024,watch,egocentric,2023,https://arxiv.org/abs/2308.05822,https://arxiv.org/abs/2308.05822,,,Language-encoded egocentric memory pipeline for storing and retrieving human-perception events,memory; episodic-memory; retrieval; assistant,egocentric-video; text; memory
EgoPoser,model,2023-08,ECCV 2024,open,egocentric,2023,https://arxiv.org/abs/2308.06493,https://siplab.org/projects/EgoPoser,,,Real-time egocentric pose estimation from sparse and intermittent HMD/controller observations,pose-and-body; xr-pose-estimation; ar-vr,head-pose; body-pose; imu
Helping Hands,model,2023-08,ICCV 2023,watch,egocentric,2023,https://arxiv.org/abs/2308.07918,https://arxiv.org/abs/2308.07918,,,Object-aware egocentric video-recognition model using hand-object cues,action-recognition; hand-object-interaction; object-recognition,egocentric-video; hand; object-boxes
Open-Vocabulary Egocentric Actions,model,2023-08,NeurIPS 2023,open,egocentric,2023,https://arxiv.org/abs/2308.11488,https://dibschat.github.io/openvocab-egoAR/,,,Open-vocabulary egocentric action-recognition framework for verb-noun action labels,open-vocabulary-hoi; action-recognition; video-language,egocentric-video; text
EgoBlur,toolkit,2023-08,arXiv,watch,egocentric,2023,https://arxiv.org/abs/2308.13093,https://arxiv.org/abs/2308.13093,,,Privacy-preserving blur pipeline and responsible-innovation analysis for Project Aria egocentric capture,wearable-sensing; tooling; ar-vr,egocentric-video; aria
POV-Surgery,dataset,2023-07,MICCAI 2023,open,egocentric,2023,https://arxiv.org/abs/2307.10387,https://batfacewayne.github.io/POV_Surgery_io/,,not specified,Egocentric surgical dataset for hand and tool pose estimation during operating-room activities,hand-object-pose; hand-object-interaction; pose-and-body,egocentric-video; hand-pose; object-pose
EgoPCA,model,2023-09,ICCV 2023,watch,egocentric,2023,https://arxiv.org/abs/2309.02423,https://arxiv.org/abs/2309.02423,,,Framework for egocentric hand-object interaction understanding using hand/object-centric perception cues,hand-object-interaction; action-recognition; object-recognition,egocentric-video; hand; object-boxes
EGOFALLS,dataset,2023-09,ICPR 2024,watch,egocentric,2023,https://arxiv.org/abs/2309.04579,https://arxiv.org/abs/2309.04579,,,Visual-audio dataset and benchmark for fall detection using egocentric cameras,activity-recognition; audio-visual; wearable-sensing,egocentric-video; audio; imu
Ego3DPose,model,2023-09,SIGGRAPH Asia 2023,watch,egocentric,2023,https://arxiv.org/abs/2309.11962,https://arxiv.org/abs/2309.11962,,,Captures 3D body cues from binocular egocentric views for pose estimation,egocentric-3d-pose; pose-and-body; ar-vr,stereo-video; body-pose; 3d-pose
SANPO,dataset,2023-09,WACV 2025,watch,egocentric,2023,https://arxiv.org/abs/2309.12172,https://arxiv.org/abs/2309.12172,,,"Scene understanding, accessibility, and human navigation dataset for egocentric navigation and spatial assistance",navigation; blind-assistance; scene-understanding; ar-sensing-navigation,egocentric-video; scene-geometry; trajectory
Egocentric Pedestrian Trajectory Benchmark,benchmark,2023-10,ICRA 2024,watch,egocentric,2023,https://arxiv.org/abs/2310.10424,https://arxiv.org/abs/2310.10424,,,Benchmarking paradigm and scale/motion-aware model for egocentric pedestrian trajectory prediction,trajectory-prediction; navigation; ar-sensing-navigation,egocentric-video; trajectory
Symbolic Active Object Localization,model,2023-10,EMNLP 2023,watch,egocentric,2023,https://arxiv.org/abs/2310.15066,https://arxiv.org/abs/2310.15066,,,Uses symbolic world knowledge to localize active objects from egocentric vision and task instructions,object-grounding; active-object-detection; spatial-reasoning; video-language,egocentric-video; text; object-boxes
IndustReal,dataset,2023-10,WACV 2024,watch,egocentric,2023,https://arxiv.org/abs/2310.17323,https://arxiv.org/abs/2310.17323,,,Egocentric industrial-like procedure-step recognition dataset with execution errors,procedure-learning; mistake-detection; action-recognition,egocentric-video
Aria-NeRF,model,2023-11,arXiv,watch,egocentric,2023,https://arxiv.org/abs/2311.06455,https://arxiv.org/abs/2311.06455,,,Multimodal egocentric view-synthesis pipeline for Project Aria-style capture using NeRF-inspired scene models,novel-view-synthesis; scene-reconstruction; ar-vr,egocentric-video; aria; scene-geometry; camera-pose
Egocentric Whole-Body MoCap,model,2023-11,CVPR 2024,watch,egocentric,2023,https://arxiv.org/abs/2311.16495,https://arxiv.org/abs/2311.16495,,,FisheyeViT plus diffusion refinement for egocentric whole-body motion capture,pose-and-body; egocentric-3d-pose; human-motion,fisheye-video; body-pose; human-motion
Exo2EgoDVC,model,2023-11,WACV 2025,open,egocentric,2023,https://arxiv.org/abs/2311.16444,https://tkhkaeio.github.io/projects/25-egodvc/,,,Dense video captioning for egocentric procedural activities using web instructional videos as exocentric supervision,captioning; procedure-learning; video-language; cross-view,egocentric-video; third-person-video; text
PALM,model,2023-11,ECCV 2024,watch,egocentric,2023,https://arxiv.org/abs/2311.17944,https://arxiv.org/abs/2311.17944,,,Predicts future actions through language models for long-term action anticipation,anticipation; action-and-procedure; planning; video-language,egocentric-video; text
LEAP,model,2023-11,arXiv,watch,egocentric,2023,https://arxiv.org/abs/2312.00055,https://arxiv.org/abs/2312.00055,,,"LLM generation of egocentric action programs with sub-actions, preconditions, postconditions, and object references",planning; procedure-learning; embodied-reasoning; video-language,egocentric-video; text
Action Scene Graphs,model,2023-12,CVPR 2024,watch,egocentric,2023,https://arxiv.org/abs/2312.03391,https://arxiv.org/abs/2312.03391,,,Temporally evolving action scene graphs for long-form egocentric video understanding,scene-graph-qa; relation-reasoning; long-context; temporal-reasoning,egocentric-video; scene-graphs
LEGO,model,2023-12,ECCV 2024,watch,egocentric,2023,https://arxiv.org/abs/2312.03849,https://arxiv.org/abs/2312.03849,,,Visual-instruction-tuned model for egocentric action-frame generation,video-generation; action-and-procedure; video-language,egocentric-video; text
LifelongMemory,model,2023-12,arXiv,watch,egocentric,2023,https://arxiv.org/abs/2312.05269,https://arxiv.org/abs/2312.05269,,,LLM-based long-form egocentric memory system for answering queries over a camera wearer's past,long-context-qa; episodic-memory; memory; assistant,egocentric-video; text; memory
EgoPlan-Bench,benchmark,2023-12,IJCV,open,egocentric,2023,https://arxiv.org/abs/2312.06722,https://github.com/ChenYi99/EgoPlan,,,Planning benchmark for multimodal large language models over egocentric human-level planning tasks,planning; embodied-reasoning; video-qa; evaluation,egocentric-video; text
CaptainCook4D,dataset,2023-12,NeurIPS 2024 Datasets and Benchmarks,watch,egocentric,2023,https://arxiv.org/abs/2312.14556,https://arxiv.org/abs/2312.14556,,,"Procedural-activity dataset for understanding execution errors, used later in multimodal temporal-action-segmentation benchmarks",mistake-detection; procedure-learning; action-recognition,egocentric-video; audio; gaze; imu
Get a Grip,model,2023-12,arXiv,open,egocentric,2023,https://arxiv.org/abs/2312.15719,https://zhifanzhu.github.io/getagrip,,,Reconstructs stable hand-object grasps from egocentric video,hand-object-pose; grasping; hand-object-interaction,egocentric-video; hand-pose; object-meshes
Rolling-Unrolling LSTMs,model,2019-05,ICCV 2019 Oral,watch,egocentric,2019,https://arxiv.org/abs/1905.09035,https://arxiv.org/abs/1905.09035,,,Multi-scale Rolling-Unrolling LSTM architecture with modality attention for anticipating egocentric actions and interacted objects,anticipation; action-and-procedure; hand-object-interaction,egocentric-video
Few-Shot First-Person Action Recognition,model,2019-07,T-PAMI,watch,egocentric,2019,https://arxiv.org/abs/1907.09382,https://arxiv.org/abs/1907.09382,,,Domain-specific priors and meta-learning approach for few-shot transfer in first-person action recognition,action-recognition; domain-adaptation; representation-learning,egocentric-video
EPIC-Fusion,model,2019-08,ICCV 2019,watch,egocentric,2019,https://arxiv.org/abs/1908.08498,https://arxiv.org/abs/1908.08498,,,"Audio-visual temporal-binding model for egocentric action recognition over RGB, optical flow, and audio",audio-visual; action-recognition; multimodal,egocentric-video; audio
Seeing and Hearing Egocentric Actions,model,2019-10,ICCV 2019 EPIC Workshop,watch,egocentric,2019,https://arxiv.org/abs/1910.06693,https://arxiv.org/abs/1910.06693,,,Study and baseline for combining audio and visual cues in egocentric human-object action understanding,audio-visual; action-recognition; hand-object-interaction,egocentric-video; audio
Motor Attention HOI Forecasting,model,2019-11,arXiv,watch,egocentric,2019,https://arxiv.org/abs/1911.10967,https://arxiv.org/abs/1911.10967,,,Joint prediction of motor attention and future actions for first-person human-object interaction forecasting,anticipation; hand-object-interaction; action-recognition,egocentric-video; hand-trajectories; object-boxes
Symbiotic Attention,model,2020-02,AAAI 2020 Oral,watch,egocentric,2020,https://arxiv.org/abs/2002.03137,https://arxiv.org/abs/2002.03137,,,Egocentric action-recognition model using privileged hand/object information through symbiotic attention,action-recognition; hand-object-interaction; multimodal,egocentric-video; hand; object-boxes
First-Person Motion-Appearance SSL,model,2020-02,ICPR 2020,watch,egocentric,2020,https://arxiv.org/abs/2002.03982,https://arxiv.org/abs/2002.03982,,,Self-supervised joint encoding of motion and appearance for first-person action recognition,action-recognition; representation-learning,egocentric-video
Unifying Few- and Zero-Shot Ego Action Recognition,model,2020-05,CVPR 2020 EPIC Workshop,watch,egocentric,2020,https://arxiv.org/abs/2006.11393,https://arxiv.org/abs/2006.11393,,,Few- and zero-shot egocentric action-recognition framework for compositional EPIC-KITCHENS-style action labels,action-recognition; open-vocabulary-hoi; video-language,egocentric-video; text
Ego-OMG,model,2020-06,arXiv,watch,egocentric,2020,https://arxiv.org/abs/2006.03201,https://arxiv.org/abs/2006.03201,,,Egocentric Object Manipulation Graph representation for activity modeling and near-future action anticipation,anticipation; hand-object-interaction; relation-reasoning,egocentric-video; object-boxes; hand
SelfPose,model,2020-11,T-PAMI 2020,watch,egocentric,2020,https://arxiv.org/abs/2011.01519,https://arxiv.org/abs/2011.01519,,,3D egocentric body-pose estimation from downward-looking headset-mounted fisheye cameras,egocentric-3d-pose; pose-and-body; ar-vr,fisheye-video; body-pose
Egocentric 4D Human Body Capture,model,2020-11,3DV 2021,watch,egocentric,2020,https://arxiv.org/abs/2011.13341,https://arxiv.org/abs/2011.13341,,,Reconstructs second-person 3D human body meshes from monocular egocentric video through 3D scene grounding,pose-and-body; scene-reconstruction; human-motion,egocentric-video; body-pose; scene-geometry
RHOI,model,2020-12,ICCV 2021,open,egocentric,2020,https://arxiv.org/abs/2012.09856,https://people.eecs.berkeley.edu/~zhecao/rhoi/,,,Reconstructs hand-object interactions in the wild without direct 3D supervision using optimization and differentiable modeling,hand-object-pose; hand-object-interaction; 3d-reconstruction,egocentric-video; hand-pose; object-meshes
Imagination-Based Ego Action Anticipation,model,2021-01,IEEE TIP 2021,watch,egocentric,2021,https://arxiv.org/abs/2101.04924,https://arxiv.org/abs/2101.04924,,,Anticipates future egocentric actions by imagining future representations before execution,anticipation; action-and-procedure; future-state-prediction,egocentric-video
Contact Representations for Action Forecasting,model,2021-02,T-PAMI 2021,watch,egocentric,2021,https://arxiv.org/abs/2102.00649,https://arxiv.org/abs/2102.00649,,,Forecasts first-person actions through contact representations built around making and breaking hand-object contact,anticipation; contact-understanding; hand-object-interaction,egocentric-video; contact; object-boxes
Indoor Future Person Localization,model,2021-03,IROS 2021,watch,egocentric,2021,https://arxiv.org/abs/2103.04019,https://arxiv.org/abs/2103.04019,,,Predicts future person locations and trajectories from egocentric wearable-camera video for assistive navigation,trajectory-prediction; navigation; blind-assistance,egocentric-video; trajectories
HPS,model,2021-03,CVPR 2021,watch,egocentric,2021,https://arxiv.org/abs/2103.17265,https://arxiv.org/abs/2103.17265,,,Human POSEitioning System for 3D human pose and self-localization in large scenes from body-mounted sensors and a head-mounted camera,egocentric-3d-pose; localization; wearable-sensing,egocentric-video; imu; body-pose; scene-scans
Egocentric Pose from Human Vision Span,model,2021-04,ICCV 2021,watch,egocentric,2021,https://arxiv.org/abs/2104.05167,https://arxiv.org/abs/2104.05167,,,Egocentric body-pose estimation from wider human-vision-span camera layouts for AR/VR,egocentric-3d-pose; pose-and-body; ar-vr,egocentric-video; body-pose
Ego-Topo 3D Map Localization,model,2021-05,ECCV 2022,watch,egocentric,2021,https://arxiv.org/abs/2105.09544,https://arxiv.org/abs/2105.09544,,,Jointly recognizes egocentric activities and localizes where they occur on a 3D environment map,activity-recognition; localization; scene-understanding,egocentric-video; scene-geometry; camera-pose
PAL,toolkit,2021-05,CVPR 2021 EPIC Workshop,watch,egocentric,2021,https://arxiv.org/abs/2105.10735,https://arxiv.org/abs/2105.10735,,,"Wearable, personalized, privacy-preserving egocentric visual-context detection system for intelligence augmentation",wearable-sensing; interactive-assistance; assistive-systems,egocentric-video; physiological-signals
RNA Cross-Domain FPV AV Recognition,model,2021-06,arXiv,watch,egocentric,2021,https://arxiv.org/abs/2106.01689,https://arxiv.org/abs/2106.01689,,,Relative Norm Alignment method for cross-domain first-person audio-visual action recognition,audio-visual; action-recognition; domain-adaptation,egocentric-video; audio
KinPoly / Dynamics-Regulated Kinematic Policy,model,2021-06,NeurIPS 2021,open,egocentric,2021,https://arxiv.org/abs/2106.05969,https://zhengyiluo.github.io/projects/kin_poly/,,,"Object-aware egocentric pose-estimation policy combining kinematics, dynamics, and scene-object information",pose-and-body; egocentric-3d-pose; human-scene-interaction,egocentric-video; body-pose; scene-geometry
Exo-to-Ego Video Synthesis,model,2021-07,ACM MM 2021,watch,egocentric,2021,https://arxiv.org/abs/2107.03120,https://arxiv.org/abs/2107.03120,,,Generates egocentric video from exocentric video for cross-view synthesis and ego-exo understanding,video-generation; cross-view; ego-exo,egocentric-video; third-person-video
MMAC Captions,dataset,2021-09,ACM MM 2021,watch,egocentric,2021,https://arxiv.org/abs/2109.02955,https://arxiv.org/abs/2109.02955,,,Sensor-augmented egocentric-video captioning dataset and dynamic-modal-attention model built around CMU-MMAC-style multimodal activity data,captioning; video-language; sensor-fusion,egocentric-video; imu; audio; text
Temporal-Context Ego Action Recognition,model,2021-11,BMVC 2021,watch,egocentric,2021,https://arxiv.org/abs/2111.01024,https://arxiv.org/abs/2111.01024,,,Transformer-based multimodal egocentric action-recognition model that attends to surrounding temporal context,action-recognition; temporal-reasoning; multimodal,egocentric-video; audio
AssistSR,benchmark,2021-11,EMNLP 2021,watch,egocentric,2021,https://arxiv.org/abs/2111.15050,https://arxiv.org/abs/2111.15050,,,Task-oriented video segment retrieval benchmark framed around personal assistants on phones and AR glasses,retrieval; temporal-grounding; interactive-assistance,video; text
E2(GO)MOTION,model,2021-12,CVPR 2022,watch,egocentric,2021,https://arxiv.org/abs/2112.03596,https://arxiv.org/abs/2112.03596,,,Motion-augmented event-stream representation for egocentric action recognition with event-camera style signals,action-recognition; sensor-fusion; efficient-inference,egocentric-video; event-stream
EPIC-KITCHENS,dataset,2018-04,ECCV 2018,open,egocentric,2018,https://arxiv.org/abs/1804.02748,https://epic-kitchens.github.io/,,research terms,"Original EPIC-KITCHENS release: 55 hours / 11.5M frames of unscripted egocentric kitchen video from 32 kitchens, with 39.6K narrated action segments and 454K object bounding boxes",action-recognition; action-detection; anticipation; hand-object-interaction,egocentric-video; audio; narrations; object-boxes
From Third Person to First Person,dataset,2018-12,CVPR 2019,watch,egocentric,2018,https://arxiv.org/abs/1812.00104,https://arxiv.org/abs/1812.00104,,,"Dataset and baselines for exocentric-to-egocentric video synthesis and retrieval, bridging third-person social-media video with first-person views",video-generation; cross-view; retrieval; ego-exo,egocentric-video; third-person-video
EgoYouTubeHands / HandOverFace,dataset,2018-03,CVPR 2018,watch,egocentric,2018,https://arxiv.org/abs/1803.03317,https://arxiv.org/abs/1803.03317,,,"Egocentric in-the-wild hand-segmentation study built around EgoYouTubeHands, HandOverFace, EGTEA, and EgoHands-style supervision",hand-segmentation; hand-detection; detection-segmentation,egocentric-video; hand-masks
Task-Dependent Gaze Transition,model,2018-03,ECCV 2018,watch,egocentric,2018,https://arxiv.org/abs/1803.09125,https://arxiv.org/abs/1803.09125,,,Egocentric gaze-prediction model that learns task-dependent attention transitions for first-person activity video,gaze-prediction; attention-modeling; action-recognition,egocentric-video; gaze
Visual Context Event Segmentation,model,2018-08,ACM MM 2018,watch,egocentric,2018,https://arxiv.org/abs/1808.02289,https://arxiv.org/abs/1808.02289,,,"Unsupervised segmentation of egocentric photostreams into contextual events, reporting a 1.5M-image lifelogging dataset and EDUB-Seg annotations",event-segmentation; lifelogging; temporal-segmentation,egocentric-images; lifelogging-images
First-Person Stories,dataset,2017-07,ICIAP 2017 Workshop,watch,egocentric,2017,https://arxiv.org/abs/1707.07863,https://arxiv.org/abs/1707.07863,,,"Egocentric first-person photo-stream dataset for lifestyle analysis, with 45K+ images labeled around eating, socializing, sedentary behavior, and other daily contexts",lifelogging; activity-recognition; lifestyle-analysis,egocentric-images; lifelogging-images
Future Person Localization,model,2017-11,CVPR 2018,watch,egocentric,2017,https://arxiv.org/abs/1711.11217,https://arxiv.org/abs/1711.11217,,,First-person wearable-camera method for localizing where people will appear in the future,trajectory-prediction; anticipation; blind-assistance,egocentric-video; trajectories
Ego2Top,dataset,2016-07,ECCV 2016,watch,egocentric,2016,https://arxiv.org/abs/1607.06986,https://arxiv.org/abs/1607.06986,,,"Ego2Top matching dataset: 50 top-view videos, 188 egocentric videos, 166K frames, and 100K body detections for matching camera wearers across ego/top views",cross-view; person-localization; tracking,egocentric-video; overhead-video; object-boxes
Ego-Engagement,dataset,2016-04,ECCV 2016,watch,egocentric,2016,https://arxiv.org/abs/1604.00906,https://arxiv.org/abs/1604.00906,,,Egocentric video dataset and model for detecting whether the camera wearer is engaged with people and objects in the physical world,engagement-detection; attention-modeling; detection-segmentation,egocentric-video; object-boxes
First-Person Activity Forecasting,model,2016-12,ICCV 2017 Oral,open,egocentric,2016,https://arxiv.org/abs/1605.09070,https://arxiv.org/abs/1612.07796,https://darko.cs.utexas.edu/,,DARKO online inverse-reinforcement-learning system for forecasting actions and active objects in first-person video,anticipation; action-recognition; active-object-detection,egocentric-video; object-boxes
Ego-Object Discovery / EDUB,dataset,2015-04,arXiv,open,egocentric,2015,https://arxiv.org/abs/1504.01639,https://arxiv.org/abs/1504.01639,,unknown-public,"EDUB egocentric photo-stream dataset for unsupervised object discovery, with 4,912 daily-life images from four users and public code/data reported by the paper",object-discovery; detection-segmentation; lifelogging,egocentric-images; object-boxes
Egocentric FOV Localization,model,2015-01,WACV 2015,watch,egocentric,2015,https://arxiv.org/abs/1501.01643,https://arxiv.org/abs/1501.01643,,,Localizes the first-person camera wearer's field of view in surveillance/top-view video using wearable point-of-view devices,fov-localization; cross-view; person-localization,egocentric-video; overhead-video
First-Person Pose Recognition,model,2015-06,CVPR 2015,watch,egocentric,2015,https://openaccess.thecvf.com/content_cvpr_2015/html/Rogez_First-Person_Pose_Recognition_2015_CVPR_paper.html,https://openaccess.thecvf.com/content_cvpr_2015/html/Rogez_First-Person_Pose_Recognition_2015_CVPR_paper.html,,,Egocentric workspace method for recognizing human pose from first-person camera geometry and likely body configurations,pose-and-body; egocentric-3d-pose; wearable-sensing,egocentric-video; body-pose
EgoPack / Backpack Full of Skills,model,2024-03,CVPR 2024,watch,egocentric,2024,https://arxiv.org/abs/2403.03037,https://arxiv.org/abs/2403.03037,,,Egocentric video understanding framework carrying multiple task perspectives across downstream Ego4D-style skills,multitask-learning; representation-learning; video-language; action-recognition,egocentric-video; text
EMAG,model,2024-05,ECCV 2024 HANDS Workshop,watch,egocentric,2024,https://arxiv.org/abs/2405.20030,https://arxiv.org/abs/2405.20030,,,Ego-motion-aware and generalizable 2D hand-forecasting method for egocentric videos,hand-forecasting; anticipation; hand-object-interaction,egocentric-video; hand-trajectories; camera-motion
Object-Aware Egocentric Online Action Detection,model,2024-06,CVPR 2024 EgoVis Workshop,watch,egocentric,2024,https://arxiv.org/abs/2406.01079,https://arxiv.org/abs/2406.01079,,,"Object-aware module for streaming online action detection in first-person footage, validated on EPIC-KITCHENS-100",online-action-detection; action-recognition; hand-object-interaction,egocentric-video; object-boxes
AFF-ttention / STAformer,model,2024-06,ECCV 2024,watch,egocentric,2024,https://arxiv.org/abs/2406.01194,https://arxiv.org/abs/2406.01194,,,Affordance- and attention-aware short-term object-interaction anticipation model for egocentric video,short-term-anticipation; affordance; hand-object-interaction,egocentric-video; object-boxes
EFM3D,benchmark,2024-06,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2406.10224,https://arxiv.org/abs/2406.10224,,,"Benchmark for 3D egocentric foundation models on Project Aria data, covering 3D object detection and surface regression",3d-object-detection; surface-regression; ar-vr; scene-understanding,aria; egocentric-video; 3d; point-clouds; camera-poses
PARSE-Ego4D,dataset,2024-06,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2407.09503,https://arxiv.org/abs/2407.09503,,,"Personal action recommendation annotations for Ego4D, including 18K+ LLM-generated and human-evaluated action suggestions",action-recommendation; interactive-assistance; video-language,egocentric-video; text; annotations
Ego-VPA,model,2024-07,WACV 2025,watch,egocentric,2024,https://arxiv.org/abs/2407.19520,https://arxiv.org/abs/2407.19520,,,Parameter-efficient adaptation method for egocentric video-language foundation models with sparse basis prompts,video-language; parameter-efficient-tuning; representation-learning,egocentric-video; text
Ego3DT,model,2024-10,ACM MM 2024,watch,egocentric,2024,https://arxiv.org/abs/2410.08530,https://arxiv.org/abs/2410.08530,,,Zero-shot 3D reconstruction and tracking framework for all objects in ego-centric videos,tracking; 3d-reconstruction; object-detection,egocentric-video; object-masks; 3d
EgoMimic,model,2024-10,ICRA 2025,watch,egocentric,2024,https://arxiv.org/abs/2410.24221,https://arxiv.org/abs/2410.24221,,,Full-stack imitation-learning framework using Project Aria egocentric human videos with 3D hand tracking plus robot data,imitation-learning; robot-learning; manipulation; cross-embodiment-transfer,aria; egocentric-video; hand-pose; robot-actions
ESOM / OVQ2D,model,2024-11,WACV 2026,watch,egocentric,2024,https://arxiv.org/abs/2411.16934,https://arxiv.org/abs/2411.16934,,,Online visual query localization with compact egocentric streaming object memory instead of full-video access,visual-query-localization; episodic-memory; tracking,egocentric-video; object-boxes
EgoCast,model,2024-12,WACV 2025,watch,egocentric,2024,https://arxiv.org/abs/2412.02903,https://arxiv.org/abs/2412.02903,,,"Bimodal 3D human-pose forecasting from egocentric video and proprioceptive data, evaluated on Ego-Exo4D and Aria Digital Twin",pose-and-body; future-state-prediction; ar-vr,egocentric-video; body-pose; proprioception; aria
GEM Ego-Vision World Model,model,2024-12,CVPR 2025,watch,egocentric,2024,https://arxiv.org/abs/2412.11198,https://arxiv.org/abs/2412.11198,,,"Generalizable ego-vision multimodal world model trained on 4,000+ hours with RGB/depth generation and ego-trajectory control",world-modeling; video-generation; future-state-prediction,egocentric-video; depth; trajectories; body-pose
EgoLog,dataset,2025-04,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2504.02624,https://arxiv.org/abs/2504.02624,,,"Fine-grained daily-log system using audio, IMU, and ubiquitous wearable signals for contextual and spatial daily activity logging",activity-recognition; lifelogging; wearable-sensing,audio; imu; wearable-sensors
Week-Long First-Person Video Privacy Probe,model,2025-04,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2504.03857,https://arxiv.org/abs/2504.03857,,,54-hour one-person first-person camera study probing how much personal-life information foundation models infer from week-long wearable video,privacy; lifelogging; personal-memory,egocentric-video; lifelogging-video; text
EgoExOR,dataset,2025-05,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2505.24287,https://arxiv.org/abs/2505.24287,,,"Ego-exo operating-room dataset with 94 minutes / 84,553 frames from wearable glasses, RGB-D cameras, ultrasound, gaze, hand tracking, audio, and scene-graph annotations",surgical-understanding; scene-graph-qa; action-recognition; hand-object-interaction,egocentric-video; exocentric-video; rgbd; gaze; audio; hand-pose; scene-graphs
GazeNLQ,model,2025-06,CVPR 2025 Ego4D Challenge,watch,egocentric,2025,https://arxiv.org/abs/2506.05782,https://arxiv.org/abs/2506.05782,,,Ego4D Natural Language Queries challenge solution that estimates gaze from video and injects it into temporal localization features,nlq; temporal-grounding; gaze,egocentric-video; gaze; text
EgoPrivacy,benchmark,2025-06,ICML 2025,watch,egocentric,2025,https://arxiv.org/abs/2506.12258,https://arxiv.org/abs/2506.12258,,,"Benchmark for privacy leakage from first-person cameras across demographic, individual, and situational inference tasks",privacy; retrieval; egocentric-vqa,egocentric-video; text
ProMemAssist,model,2025-07,UIST 2025,watch,egocentric,2025,https://arxiv.org/abs/2507.21378,https://arxiv.org/abs/2507.21378,,,Smart-glasses proactive-assistance system that models working memory from multimodal wearable signals to time interventions,proactive-assistance; memory; wearable-sensing,smart-glasses; wearable-sensors; egocentric-video
EgoPrompt,model,2025-08,ACM MM 2025,watch,egocentric,2025,https://arxiv.org/abs/2508.03266,https://arxiv.org/abs/2508.03266,,,Prompt-learning framework for egocentric action recognition that models verb-noun semantic and contextual relationships,action-recognition; video-language; representation-learning,egocentric-video; text
EgoTwin,model,2025-08,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2508.13013,https://arxiv.org/abs/2508.13013,,,Joint egocentric video and human-motion generation framework that aligns first-person view dynamics with head-centric motion,video-generation; human-motion; world-modeling,egocentric-video; human-motion; trajectories
EvHand-FPV,dataset,2025-09,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2509.13883,https://arxiv.org/abs/2509.13883,,,Event-based first-person-view 3D hand-tracking dataset and lightweight ROI-based tracking framework,3d-hand-pose; hand-tracking; efficient-inference,event-stream; egocentric-video; hand-pose
ActiveUMI,toolkit,2025-10,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2510.01607,https://arxiv.org/abs/2510.01607,,,Portable VR/UMI-style collection system capturing active egocentric perception from head-mounted displays for bimanual robot manipulation,robot-learning; imitation-learning; active-perception; bimanual-manipulation,egocentric-video; wrist-camera-video; robot-actions; head-pose
egoEMOTION,dataset,2025-10,NeurIPS 2025,watch,egocentric,2025,https://arxiv.org/abs/2510.22129,https://arxiv.org/abs/2510.22129,,,"50+ hours from 43 participants with Project Aria egocentric video, eye tracking, PPG, IMU, emotion self-reports, and Big Five personality labels",affect-recognition; personality-recognition; wearable-sensing,aria; egocentric-video; gaze; physiological-signals; imu
WAGIBench,benchmark,2025-10,NeurIPS 2025 Spotlight,watch,egocentric,2025,https://arxiv.org/abs/2510.22443,https://arxiv.org/abs/2510.22443,,,"Assistive wearable-agent goal-inference benchmark with 29 hours, 348 participants, and 3,477 multimodal recordings",goal-inference; situated-assistance; multimodal-reasoning,egocentric-video; audio; digital-context; wearable-sensors
Eyes on Target,model,2025-11,RAAI 2025,watch,egocentric,2025,https://arxiv.org/abs/2511.01237,https://arxiv.org/abs/2511.01237,,,"Depth-aware gaze-guided object detection framework for egocentric video, evaluated on simulator data and Ego4D/Ego-CH-Gaze-style benchmarks",object-detection; gaze; attention-modeling,egocentric-video; gaze; depth; object-boxes
SkillSight,model,2025-11,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2511.19629,https://arxiv.org/abs/2511.19629,,,"Power-efficient first-person skill assessment framework that distills video+gaze teacher models into a gaze-only student across cooking, music, and sports",skill-assessment; gaze; efficient-inference,egocentric-video; gaze
Mistake Attribution / MATT,benchmark,2025-11,CVPR 2026,watch,egocentric,2025,https://arxiv.org/abs/2511.20525,https://arxiv.org/abs/2511.20525,,,"Fine-grained mistake-attribution task and MisEngine data engine producing EPIC-KITCHENS-M and Ego4D-M annotations for semantic, temporal, and spatial mistake understanding",mistake-detection; mistake-attribution; procedural-reasoning,egocentric-video; annotations; object-boxes
X-Humanoid,model,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.04537,https://arxiv.org/abs/2512.04537,,,Human-to-humanoid video translation pipeline applying a trained model to 60 hours of Ego-Exo4D and releasing 3.6M robotized humanoid video frames,video-generation; humanoid-control; world-modeling,egocentric-video; exocentric-video; humanoid-motion; synthetic-video
DreamTacVLA,model,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.23864,https://arxiv.org/abs/2512.23864,,,Contact-rich VLA framework combining high-resolution tactile images with wrist-camera local vision and third-person macro vision,vla; tactile-learning; contact-understanding; manipulation,wrist-camera-video; tactile; exocentric-video; robot-actions
EgoGrasp,model,2026-01,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2601.01050,https://arxiv.org/abs/2601.01050,,,"World-space hand-object interaction reconstruction from dynamic egoview videos, supporting open-vocabulary objects",hand-object-pose; 3d-reconstruction; open-vocabulary-hoi,egocentric-video; hand-pose; object-pose; 3d
EgoHandICL,model,2026-01,ICLR 2026,watch,egocentric,2026,https://arxiv.org/abs/2601.19850,https://arxiv.org/abs/2601.19850,https://github.com/Nicous20/EgoHandICL,,First in-context learning method for robust egocentric 3D hand reconstruction using VLM-retrieved exemplars and multimodal context under occlusion,3d-hand-reconstruction; egocentric-vision; hand-object-interaction; representation-learning,egocentric-video; hand-pose; hand-mesh; 3d; model
EgoActor,model,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.04515,https://arxiv.org/abs/2602.04515,,,"VLM for grounding high-level instructions into spatially aware egocentric humanoid actions including locomotion, head movement, manipulation, and HRI commands",humanoid-control; vla; spatial-reasoning; manipulation,egocentric-video; text-instructions; robot-actions; humanoid-motion
AirGlove,benchmark,2026-02,ICASSP 2026,watch,egocentric,2026,https://arxiv.org/abs/2602.05159,https://arxiv.org/abs/2602.05159,,,Evaluation and dataset for egocentric 3D hand tracking of gloved hands under sensing-glove appearance shifts,3d-hand-pose; hand-tracking; wearable-sensing,egocentric-video; hand-pose; glove-sensors
EgoHumanoid,model,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.10106,https://arxiv.org/abs/2602.10106,,,Humanoid loco-manipulation framework co-training VLA policies on robot-free egocentric human demonstrations and limited robot data,humanoid-control; vla; imitation-learning; loco-manipulation,egocentric-video; human-demonstrations; robot-actions; humanoid-motion
WHOLE,model,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.22209,https://arxiv.org/abs/2602.22209,,,World-grounded hand-object reconstruction from egocentric videos using a generative prior over hand-object motion,hand-object-pose; 3d-reconstruction; hand-object-interaction,egocentric-video; hand-pose; object-pose; 3d
HOI-Synth,benchmark,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.29733,https://fpv-iplab.github.io/HOI-Synth/,https://fpv-iplab.github.io/HOI-Synth/,,"Synthetic egocentric hand-object interaction benchmark adding automatically labeled synthetic contact and HOI data to VISOR, EgoHOS, and ENIGMA-51",hoi-detection; synthetic-data; data-augmentation; benchmark,synthetic-data; egocentric-image; hand-object-interaction; segmentation; bounding-boxes
EgoAVFlow,model,2026-02,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2602.22461,https://arxiv.org/abs/2602.22461,,,Robot policy learning from egocentric human videos through shared 3D flow for manipulation and active camera/viewpoint control,robot-learning; active-perception; imitation-learning; manipulation,egocentric-video; 3d-flow; robot-actions; camera-trajectory
Pandora Articulated 3D Scene Graphs,model,2026-03,BMVC 2025,watch,egocentric,2026,https://arxiv.org/abs/2603.28732,https://arxiv.org/abs/2603.28732,,,"Articulated 3D scene-graph recovery from human Project Aria egocentric exploration, transferring object articulation knowledge to robots",scene-graph-qa; 3d-scene-understanding; robot-learning; articulated-objects,aria; egocentric-video; scene-graphs; 3d
RoSHI,toolkit,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.07331,https://arxiv.org/abs/2604.07331,,,Robot-oriented wearable suit fusing sparse IMUs with Project Aria to estimate global 3D pose and body shape for in-the-wild human data,pose-and-body; robot-learning; human-motion; data-collection,aria; imu; egocentric-video; body-pose; human-demonstrations
VISTA Daily Assistance,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.10579,https://arxiv.org/abs/2605.10579,,,Generative egocentric-video framework that synthesizes reactive and proactive daily-assistance scenarios for agent training and evaluation,video-generation; proactive-assistance; world-modeling,egocentric-video; synthetic-video; text
Low-Resolution Active Perception BC,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.14106,https://arxiv.org/abs/2605.14106,,,Behavior-cloning study showing a wrist-mounted low-resolution egocentric RGB camera can learn active perception for object finding and grasp triggering,active-perception; robot-learning; manipulation,wrist-camera-video; robot-actions
DeMiAn,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.17077,https://arxiv.org/abs/2605.17077,,,Dense multi-aspect language re-annotation pipeline improving VLA policies and world-action models over 1M robot clips and 50K EgoVerse human-egocentric videos,vla; robot-learning; language-grounding; manipulation,egocentric-video; language-instructions; robot-actions
Egocentric Nursing Competency Assessment,model,2026-05,CVPR 2026 Workshop,watch,egocentric,2026,https://arxiv.org/abs/2605.20233,https://arxiv.org/abs/2605.20233,,,3.8 hours of egocentric nursing-simulation video with 493 annotated actions used to relate predicted action timelines to instructor-rated competency,skill-assessment; action-recognition; medical-assistance,egocentric-video; action-labels
Gaze2Act,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.30282,https://arxiv.org/abs/2605.30282,,,Gaze-conditioned VLA framework mapping first-person gaze into robot perspective for target specification in interactive manipulation,vla; gaze; manipulation; robot-learning,egocentric-video; gaze; robot-actions
Qwen-VLA,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.30280,https://arxiv.org/abs/2605.30280,,,"Unified embodied foundation model trained across manipulation trajectories, egocentric human demonstrations, simulation, navigation, and trajectory supervision",vla; robot-learning; navigation; manipulation,egocentric-video; robot-actions; language-instructions; trajectories
CAIP,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.17256,https://arxiv.org/abs/2606.17256,,,"Contrastive action-image pretraining using 3D hand keypoints from 32,041 hours of egocentric human video plus 88 hours of robot manipulation data",representation-learning; robot-learning; manipulation; vla,egocentric-video; hand-pose; robot-actions
Qwen-RobotManip,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.17846,https://arxiv.org/abs/2606.17846,,,"Qwen-VL-based manipulation foundation model with a 38,100-hour pretraining corpus that converts egocentric hand demonstrations into robot trajectories across 15 platforms",vla; robot-learning; cross-embodiment-transfer; manipulation,egocentric-video; hand-pose; robot-actions; language-instructions
PAIWorld,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.18375,https://arxiv.org/abs/2606.18375,,,"3D-consistent multi-view world foundation model for robotic manipulation across egocentric, eye-to-hand, and wrist-mounted observations",world-modeling; robot-learning; video-generation; manipulation,egocentric-video; wrist-camera-video; multiview-video; 3d
Mem-World,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.18960,https://arxiv.org/abs/2606.18960,,,Memory-augmented action-conditioned world model with a wrist-view-centered 4D surfel memory for persistent manipulation rollouts,world-modeling; robot-learning; manipulation; memory,wrist-camera-video; robot-actions; 3d
Pose6DAug,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.20118,https://arxiv.org/abs/2606.20118,,,Physically plausible multi-view object-swapping augmentation for VLA robot manipulation policies under heavy occlusion and egocentric viewpoints,data-augmentation; vla; manipulation; robot-learning,egocentric-video; multiview-video; robot-actions; object-pose
EgoEvGesture,dataset,2025-03,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2503.12419,https://arxiv.org/abs/2503.12419,,,"First large-scale egocentric event-camera gesture-recognition dataset, paired with a 7M-parameter head-motion-robust network (62.7% on unseen subjects, 97.0% on DVS128 Gesture)",pose-and-body; hand-object-interaction; ar-sensing-navigation,event-camera; egocentric-video
EgoSpanLift,model,2025-11,NeurIPS 2025 spotlight,watch,egocentric,2025,https://arxiv.org/abs/2511.18470,https://arxiv.org/abs/2511.18470,,,"Lifts 2D egocentric gaze forecasting into 3D visual-span prediction with SLAM keypoints, volumetric regions, and a 3D U-Net plus transformer, over a 364.6K-sample benchmark (Gaze Beyond the Frame)",ar-sensing-navigation; pose-and-body; anticipation,egocentric-video; gaze; camera-pose
PRISM,dataset,2026-03,arXiv,request,egocentric,2026,https://arxiv.org/abs/2603.29281,https://dreamvu.ai/prism,,cc-by-nc-4.0,"Paper reports a 270K-sample multi-view retail SFT corpus; the public gated PRISM-100K subset contains 100K samples, 25,173 videos (10,623 ego and 14,550 exo), 21 task types, and about 772 GB",embodied-ai; spatial-reasoning; physical-action-understanding; question-answering; representation-learning,egocentric-video; exocentric-video; first-person-360-video; text; qa
TacThru-UMI,toolkit,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.09851,https://arxiv.org/abs/2512.09851,,,See-through-skin tactile-visual sensor and UMI-style imitation-learning framework for manipulation with simultaneous visual perception and robust tactile tracking,vla; tactile-learning; robot-learning; manipulation; imitation-learning,wrist-camera-video; tactile; robot-actions; human-demonstrations
FARM,model,2025-10,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2510.13324,https://tactile-farm.github.io,https://tactile-farm.github.io,,"Force-aware diffusion-policy framework using a modified UMI gripper with GelSight Mini tactile sensing and an actuated deployment gripper for high-, low-, and dynamic-force tasks",robot-learning; tactile-learning; manipulation; imitation-learning; dexterous-manipulation,wrist-camera-video; tactile; force; robot-actions; human-demonstrations
UMI on Legs,model,2024-07,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2407.10353,https://umi-on-legs.github.io,https://umi-on-legs.github.io,,Combines handheld UMI task demonstrations with simulation-trained whole-body controllers to transfer manipulation policies onto quadruped embodiments,robot-learning; cross-embodiment-transfer; mobile-manipulation; imitation-learning; whole-body-manipulation,wrist-camera-video; robot-actions; human-demonstrations; trajectories
UMI-on-Air,model,2025-10,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2510.02614,https://arxiv.org/abs/2510.02614,,,Embodiment-aware diffusion policy that adapts handheld UMI demonstrations to aerial manipulators through controller-feedback-guided sampling,robot-learning; cross-embodiment-transfer; manipulation; mobile-manipulation; policy-optimization,wrist-camera-video; robot-actions; human-demonstrations; trajectories
EmbodiSwap,model,2025-10,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2510.03706,https://arxiv.org/abs/2510.03706,,,"Photorealistic robot-overlay generation over in-the-wild ego-centric human videos, with synthetic robot datasets over EPIC-KITCHENS, HOI4D, and Ego4D for zero-shot imitation learning",robot-learning; cross-embodiment-transfer; video-generation; imitation-learning; manipulation,egocentric-video; synthetic-video; robot-actions; human-demonstrations
Mitty,model,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.17253,https://arxiv.org/abs/2512.17253,,,Diffusion Transformer for human-to-robot video generation that synthesizes paired human-robot examples from large egocentric datasets and evaluates on EPIC-KITCHENS,video-generation; robot-learning; cross-embodiment-transfer; world-modeling,egocentric-video; synthetic-video; robot-actions; human-demonstrations
tau0-WM,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.01027,https://arxiv.org/abs/2606.01027,,,"Unified video-action world model trained on about 27,300 hours of real-robot teleoperation, UMI-style interaction, egocentric human videos, and rollout/failure trajectories",world-modeling; robot-learning; vla; manipulation; action-prediction,wrist-camera-video; egocentric-video; multiview-video; robot-actions; language-instructions
DR-MV3D,model,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.23557,https://arxiv.org/abs/2606.23557,,,"Dense-reward framework for multi-view 3D VQA that decomposes reasoning into global-map construction, question-conditioned view planning, and egocentric answer grounding",spatial-reasoning; video-qa; 3d-scene-understanding; embodied-reasoning; reward-modeling,egocentric-video; multiview-video; 3d; qa; text
EventPrune,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.19506,https://arxiv.org/abs/2605.19506,,,"Training-free event-camera-guided visual-token pruning for first-person dynamic spatial reasoning, introducing the ESR-Real RGB-event benchmark",efficient-inference; spatial-reasoning; video-qa; ar-sensing-navigation,first-person-video; event-camera; qa
SpatioRoute,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.18209,https://arxiv.org/abs/2605.18209,,,"Question-aware prompt routing for zero-shot egocentric spatial QA, specializing prompts by question type without fine-tuning or 3D sensor input",spatial-reasoning; video-qa; zero-shot; prompt-learning,egocentric-video; qa; text
MAGIC-Video,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.08271,https://github.com/lijiazheng0917/MAGIC-video,https://github.com/lijiazheng0917/MAGIC-video,,"Training-free multimodal memory graph and narrative-chain framework for ultra-long agentic video reasoning over egocentric recordings, live streams, and surveillance footage",memory; long-video-reasoning; video-qa; agentic-reasoning; retrieval,egocentric-video; text; qa; scene-graphs
World2VLM,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.26934,https://arxiv.org/abs/2604.26934,,,Training-time distillation from view-consistent world-model imagination into VLMs for dynamic spatial reasoning under egocentric motion,world-modeling; spatial-reasoning; video-language; representation-learning,egocentric-video; synthetic-video; camera-trajectory; text
Video ReCap,model,2024-02,CVPR 2024,watch,egocentric,2024,https://arxiv.org/abs/2402.13250,https://sites.google.com/view/vidrecap,https://sites.google.com/view/vidrecap,,"Recursive captioning model for videos from one second to two hours, plus Ego4D-HCap with 8,267 long-range summaries for hour-long egocentric videos",captioning; long-video-understanding; video-qa; memory,egocentric-video; captions; text
GPT4Ego,model,2024-01,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2401.10039,https://arxiv.org/abs/2401.10039,,,"Zero-shot egocentric action-recognition framework that aligns fine-grained concept descriptions with VLM features over EPIC-KITCHENS-100, EGTEA, and Charades-Ego",zero-shot; action-recognition; video-language; representation-learning,egocentric-video; text
EVA02-AT,model,2025-06,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2506.14356,https://github.com/xqwang14/EVA02-AT,https://github.com/xqwang14/EVA02-AT,,EVA02-based egocentric video-language foundation models with spatial-temporal rotary positional embeddings and symmetric multi-similarity optimization,video-language-pretraining; video-text-retrieval; representation-learning; efficient-inference,egocentric-video; text
X-MIC,model,2024-03,ECCV 2024,watch,egocentric,2024,https://arxiv.org/abs/2403.19811,https://github.com/annusha/xmic,https://github.com/annusha/xmic,,"Cross-modal instance conditioning framework for adapting VLMs to egocentric action generalization across EPIC-KITCHENS, Ego4D, and EGTEA",action-recognition; domain-generalization; video-language; representation-learning,egocentric-video; text
EgoAdapt Efficient Perception,model,2025-06,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2506.21080,https://arxiv.org/abs/2506.21080,,,"Adaptive multisensory distillation and policy learning for efficient egocentric action recognition, active-speaker localization, and behavior anticipation on EPIC-KITCHENS, EasyCom, and AEA",efficient-inference; action-recognition; audio-visual; anticipation; ar-sensing-navigation,egocentric-video; audio; gaze; imu
Spatial Cognition / LMK,model,2024-04,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2404.05072,https://arxiv.org/abs/2404.05072,,,"Out-of-sight active-object tracking from egocentric video using Lift, Match, and Keep over 100 long EPIC-KITCHENS videos",spatial-reasoning; object-tracking; memory; 3d-scene-understanding,egocentric-video; 3d; object-tracks
TimeSearch-R,model,2025-11,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2511.05489,https://github.com/Time-Search/TimeSearch-R,https://github.com/Time-Search/TimeSearch-R,,"Reinforcement-learned temporal search for long-form video understanding with completeness self-verification, evaluated on Haystack-Ego4D and broader long-video benchmarks",long-video-understanding; temporal-retrieval; video-qa; reinforcement-learning,egocentric-video; text; qa
Know-Show,benchmark,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.05513,https://github.com/LUNAProject22/Know-Show,https://github.com/LUNAProject22/Know-Show,,"2.5K human-authored questions over Charades, Action Genome, and Ego4D that jointly evaluate spatio-temporal reasoning and visual/temporal grounding",video-qa; spatial-reasoning; temporal-grounding; grounded-reasoning; benchmark,egocentric-video; video; qa; annotations
PCSR-Bench,benchmark,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.12413,https://arxiv.org/abs/2605.12413,,,"84,373 QA pairs from 2,600 omnidirectional indoor images over eight perspective-conditioned spatial reasoning tasks including egocentric rotation and limited-FOV visibility",spatial-reasoning; question-answering; embodied-reasoning; benchmark,first-person-360-video; egocentric-image; qa; annotations
SpaMEM,benchmark,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.22409,https://huggingface.co/datasets/mill-ct-liao/SpaMEM,,,"Dynamic spatial-memory benchmark with 10.6M high-fidelity images across RGB/depth/instance/semantic modalities from 25K+ action-conditioned interaction sequences in 1,000 houses",spatial-reasoning; memory; benchmark; embodied-reasoning,egocentric-video; depth; segmentation; annotations; qa
Ego4D-OSCA,benchmark,2024-05,arXiv,watch,egocentric,2024,https://arxiv.org/abs/2405.12789,https://arxiv.org/abs/2405.12789,,,"Object-state-change anticipation annotations over Ego4D procedural video, paired with a baseline that integrates visual and language history",anticipation; object-state-change; procedural-reasoning; benchmark,egocentric-video; annotations; text
OR-Action,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.13332,https://arxiv.org/abs/2606.13332,,,"Fine-grained multi-role action benchmark built on a public ego-exocentric operating-room dataset, deriving dense actions from scene-graph state changes",surgical-understanding; action-recognition; temporal-segmentation; scene-graph-qa; benchmark,egocentric-video; exocentric-video; scene-graphs; annotations
IMPACT-HOI,toolkit,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.01666,https://github.com/541741106/IMPACT_HOI,https://github.com/541741106/IMPACT_HOI,,Mixed-initiative supervisory-control framework for constructing onset-anchored partial HOI event graphs in egocentric procedural video,annotation-tooling; hand-object-interaction; procedural-reasoning; robot-learning,egocentric-video; annotations; scene-graphs
EgoScreen-Emotion / ESE,benchmark,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.15823,https://arxiv.org/abs/2604.15823,,,"224 movie trailers captured under egocentric screen-view conditions with 28,667 temporally aligned keyframes and confidence-aware multi-label emotion annotations",affect-recognition; long-context; audio-visual; benchmark,egocentric-video; audio; annotations
EyeCue,model,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.07859,https://github.com/langzhang2000/EyeCue,https://github.com/langzhang2000/EyeCue,,"Gaze-empowered egocentric video model and CogDrive annotations for driver cognitive-distraction detection across road, weather, and time-of-day settings",driving; internal-state-reasoning; affect; safety; multimodal-reasoning,egocentric-video; gaze; annotations
GIST,toolkit,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.15495,https://arxiv.org/abs/2604.15495,,,"Grounded Intelligent Semantic Topology pipeline that turns mobile point clouds into semantic navigation topologies for search, localization, zone classification, and egocentric route instructions",navigation; assistive-systems; spatial-localization; semantic-mapping; instruction-generation,point-clouds; maps; text; egocentric-video
Gaze-to-Guidance Assistants,model,2026-04,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2604.08062,https://arxiv.org/abs/2604.08062,,,Controlled study of gaze-grounded multimodal LLM assistance using egocentric video with gaze overlays to infer reading difficulty and personalize retrospective help,assistive-systems; gaze-reasoning; personalized-qa; cognitive-state-aware-assistance,egocentric-video; gaze; text
VueBuds,toolkit,2026-03,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2603.29095,https://arxiv.org/abs/2603.29095,,,"Camera-integrated wireless earbuds for egocentric vision, streaming low-power binocular monochrome views to on-device VLMs for scene understanding, translation, visual reasoning, and text reading",ar-sensing-navigation; wearable-interaction; assistive-systems; on-device-perception,egocentric-video; monochrome; wearable-sensing; text
ROHIT,model,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.07394,https://arxiv.org/abs/2512.07394,,,"Reconstructs Objects along Hand Interaction Timelines from HOT3D and EPIC-KITCHENS stable-grasp clips, propagating object pose across pre-contact, held, and released phases",object-reconstruction; hand-object-interaction; 3d-reconstruction; tracking,egocentric-video; hand-pose; object-pose; 3d
Single-View Mesh Rotation Stress Test,benchmark,2026-06,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2606.22987,https://arxiv.org/abs/2606.22987,,,"Controlled roll/pitch/yaw stress-test protocol for single-view mesh reconstruction on Aria Digital Twin and Franka wrist-camera sequences, exposing view-dependent layout and depth failures",3d-reconstruction; digital-twin-evaluation; robot-learning; benchmark,aria; wrist-camera-video; 3d; object-pose
The N-Body Problem,model,2025-12,arXiv,watch,egocentric,2025,https://arxiv.org/abs/2512.11393,https://arxiv.org/abs/2512.11393,,,"VLM reasoning task for predicting feasible N-person parallel execution from a single egocentric video, evaluated on 100 EPIC-KITCHENS and HD-EPIC videos",procedural-reasoning; spatial-reasoning; planning; video-language,egocentric-video; text; qa
World Action Models,survey,2026-05,arXiv,watch,egocentric,2026,https://arxiv.org/abs/2605.12090,https://arxiv.org/abs/2605.12090,,,"Survey and taxonomy of world action models that unifies VLA, predictive world modeling, robot teleoperation, portable human demonstrations, simulation, and internet-scale egocentric video",world-modeling; vla; robot-learning; survey,survey; egocentric-video; robot-actions
FutureNav,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.30367,https://arxiv.org/abs/2606.30367,,,"Unified world-action modeling framework for vision-and-language navigation, jointly optimizing action prediction, inverse/forward dynamics, and future spatial-state prediction from egocentric agent observations",navigation; world-modeling; vla; spatial-reasoning; robot-learning,egocentric-video; text; spatial-features; trajectories
ZR-0,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.30552,https://github.com/RUCKBReasoning/ZR-0,https://github.com/RUCKBReasoning/ZR-0,,2.6B VLA model trained with dense embodied chain-of-thought supervision over ProcCorpus-60M to align cross-embodiment manipulation representations without inference-time reasoning overhead,vla; robot-learning; manipulation; cross-embodiment-transfer; reasoning,video; text; robot-actions; trajectories
VLK,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.30645,https://vision-language-kinematics.github.io/,,,"Vision-language-kinematics supervision pipeline generating 48K synthetic humanoid loco-manipulation trajectories with paired egocentric renderings, instructions, and whole-body kinematic targets in reconstructed scenes",humanoid-control; loco-manipulation; vla; robot-learning; synthetic-data,synthetic-video; egocentric-video; text; trajectories; humanoid-motion
HUMEMBR,model,2026-06,IROS 2026,watch,adjacent,2026,https://arxiv.org/abs/2606.30404,https://arxiv.org/abs/2606.30404,,,"Human-Centered Memory for Embodied Robots, building structured long-term human-routine memories for embodied question answering and routine-conditioned navigation on physical robots",memory-reasoning; navigation; embodied-reasoning; question-answering; robot-learning,robot-observations; memory; text; trajectories
WARP Retarget,model,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.29940,https://warp-retarget.github.io/,,,Whole-body-aware retargeting pipeline that converts offline human demonstrations into precise mobile-manipulator trajectories using shoulder-elbow-wrist geometric solving and lazy mobile-base control,robot-learning; imitation-learning; retargeting; mobile-manipulation; cross-embodiment-transfer,human-demonstrations; human-motion; trajectories; robot-actions
ContactWorld,benchmark,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.13877,https://arxiv.org/abs/2606.13877,,,"Vision-tactile world-model benchmark across 12 contact-rich robot manipulation tasks, comparing wrist-view, front-view, point-cloud, and tactile force-field representations",world-modeling; tactile-learning; robot-learning; manipulation; benchmark,wrist-camera-video; exocentric-video; tactile; point-clouds; robot-actions
ACWM-Phys,benchmark,2026-05,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2605.08567,https://arxiv.org/abs/2605.08567,,,"Action-conditioned video world-model benchmark for generalized physical interaction, spanning rigid, kinematic, deformable, and particle dynamics in a controllable simulator",world-modeling; physical-action-understanding; benchmark; evaluation,synthetic-video; robot-actions
3DG-VLN / UAV-VLN-FOV,benchmark,2026-06,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2606.20045,https://github.com/xuefanfu/3DG-VLN,https://github.com/xuefanfu/3DG-VLN,,"2,717 high-resolution target-visible UAV trajectories with front/downward egocentric observations and continuous 3D waypoint annotations for see-and-reach VLN",navigation; spatial-reasoning; benchmark; robot-learning,egocentric-video; trajectory; text; annotations
JAXenstein,benchmark,2026-05,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2605.19926,https://arxiv.org/abs/2605.19926,,,JAX-based Wolfenstein 3D renderer for accelerated reinforcement-learning benchmarks in visual first-person tasks under partial observability,reinforcement-learning; benchmark; virtual-egocentric-reasoning,first-person-video; synthetic-video
DenseStep2M,dataset,2026-04,arXiv,watch,adjacent,2026,https://arxiv.org/abs/2604.26565,https://huggingface.co/datasets/mingjige/DenseStep2M,,,"Training-free procedural-video annotation pipeline producing about 100K videos and 2M dense instructional steps, with zero-shot transfer evaluations on egocentric, exocentric, and mixed-perspective domains",procedure-understanding; captioning; temporal-grounding; retrieval; benchmark,video; captions; text
RoboPearls,toolkit,2025-06,arXiv,watch,adjacent,2025,https://arxiv.org/abs/2506.22756,https://arxiv.org/abs/2506.22756,,,"Editable 3D Gaussian video-simulation framework for robot manipulation, evaluated on RLBench, COLOSSEUM, Ego4D, Open X-Embodiment, and real-robot scenes",world-modeling; robot-learning; video-generation; manipulation,synthetic-video; robot-actions; 3d