大模型的记忆管理：从记忆外挂到主动管理
Junrulu
• • 2NEW Articles from Team or Enterprise organizations will get promoted to the main section.
Build a Tiny Language Model, Day 4: Save the Experiment
ericmey
• Leading the System One Mosaic Benchmark: What Darwin-27B-ZTC-v2's #1 Means
FINAL-Bench
• • 6Testing selected-token J-lens directions and a fixed-symbol autoencoder on Qwen3-8B
kotlarmilos
• 【观察】Opus 5.5 做视频，强在哪一层？——一个小团队的制作笔记
BreadStudio
• Understanding Jev: confidence is only useful if it works
stephen-solka
• Using Token Data to Improve the AI Agent Experience
phranzia
• • 2FEM-ASM: Modular Language Computation
Bochkov
• Introducing Superfluid: the local-AI native LLM server for multi-agent inference
basecompute
• • 3JOSIE-2 Technical Report: Curated Post-Training for Reasoning and Behavior in Small Language Models
Goekdeniz-Guelmez
• • 1LightOnOCR-3: High-Performance OCR and Layout Extraction in One Model
lightonai
• • 24GLiNER Can Now Detect Names and PII in Indian Languages, in Their Own Scripts
nishikantmandal007
• • 3funes now recalls in more than English
dacorvo
• Carbon-A: Finding genes in known and unknown genomes
HuggingFaceBio
• • 19Six Questions to Ask a LeRobot Dataset Before You Train
harpreetsahota
• • 3RLCD: Are We Moving from AI Models That Generate Answers to Models That Make Calibrated Decisions?
javadtaghia
• Claude Opus 5.5 is the new #1 on Multilingual Terminal-Bench, the coding benchmark on AURORA, our multilingual AI leaderboard.
Lilt-org
• Can one EEG encoder answer several questions?
Twu31
• Topology-Aware Hybrid Retrieval for Enterprise Knowledge Systems
shubhayjoshua
• How Speculative Decoding Accelerates LLM Token Generation
prismberry-technologies
•