Per-row records and fitted word directions for SYNC. Paper: arxiv.org/abs/2610.19696 · Code: github.com/LiXin97/SYNC
XinLi PRO
XINLI1997
AI & ML interests
None yet
Recent Activity
updated a model 2 days ago
XINLI1997/WirelessMathLM-Qwen3-4B updated a model 2 days ago
XINLI1997/WirelessMathLM-7B updated a model 2 days ago
XINLI1997/WirelessMathLM-0.5BOrganizations
When does on-policy distillation transfer teacher errors? Paper, checkpoints and data. Code: github.com/LiXin97/ErrOPD
NeurIPS 2026 E&D: WirelessMathBench-XL dataset and WirelessMathLM GRPO checkpoints.
Checkpoints and data for ReMaRk (GRPO for LLM recommenders without responder noise). Code: github.com/LiXin97/ReMaRk
-
Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders
Paper • 2610.17771 • Published • 13 -
XINLI1997/ReMaRk-data
Viewer • Updated • 878k • 285 • 1 -
XINLI1997/ReMaRk-Qwen3.5-9B-RedditV2-ReMaRk-s42
Text Generation • 9B • Updated • 292 • 1 -
XINLI1997/Llama-3.1-8B-ReMaRk-RedditV2-ReMaRk-s3407
Text Generation • 8B • Updated • 342
Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation. Data and models.
-
Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation
Paper • 2609.35347 • Published • 75 -
XINLI1997/DN-MOPD-Data
Viewer • Updated • 26.7k • 83 • 1 -
XINLI1997/DN-MOPD-Qwen3.5-9B
Image-Text-to-Text • 9B • Updated • 24 • 1 -
XINLI1997/DN-MOPD-Qwen3.5-4B
Image-Text-to-Text • 5B • Updated • 22 • 1
Per-row records and fitted word directions for SYNC. Paper: arxiv.org/abs/2610.19696 · Code: github.com/LiXin97/SYNC
Checkpoints and data for ReMaRk (GRPO for LLM recommenders without responder noise). Code: github.com/LiXin97/ReMaRk
-
Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders
Paper • 2610.17771 • Published • 13 -
XINLI1997/ReMaRk-data
Viewer • Updated • 878k • 285 • 1 -
XINLI1997/ReMaRk-Qwen3.5-9B-RedditV2-ReMaRk-s42
Text Generation • 9B • Updated • 292 • 1 -
XINLI1997/Llama-3.1-8B-ReMaRk-RedditV2-ReMaRk-s3407
Text Generation • 8B • Updated • 342
When does on-policy distillation transfer teacher errors? Paper, checkpoints and data. Code: github.com/LiXin97/ErrOPD
Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation. Data and models.
-
Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation
Paper • 2609.35347 • Published • 75 -
XINLI1997/DN-MOPD-Data
Viewer • Updated • 26.7k • 83 • 1 -
XINLI1997/DN-MOPD-Qwen3.5-9B
Image-Text-to-Text • 9B • Updated • 24 • 1 -
XINLI1997/DN-MOPD-Qwen3.5-4B
Image-Text-to-Text • 5B • Updated • 22 • 1
NeurIPS 2026 E&D: WirelessMathBench-XL dataset and WirelessMathLM GRPO checkpoints.