Rethinking Token Reweighting for SFT: Suppress, Reverse, and Extrapolate Learned Features
Paper • 2609.33463 • Published • 11
张畅
LunarDawn
·
AI & ML interests
None yet
Recent Activity
upvoted a paper about 7 hours ago
Rethinking Token Reweighting for SFT: Suppress, Reverse, and Extrapolate Learned Features upvoted a paper 16 days ago
RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning upvoted a paper 24 days ago
HyQuant: Hybrid-Precision Quantization for LLM AttentionOrganizations
None yet