AI Breakdown

Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning

Sep 24, 2025
Ask episode
Chapters
Transcript
Episode notes