Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning Paper • 2506.01939 • Published Jun 2 • 185
Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model? Paper • 2504.13837 • Published Apr 18 • 136
Model Surgery: Modulating LLM's Behavior Via Simple Parameter Editing Paper • 2407.08770 • Published Jul 11, 2024 • 21
