GMTS: Gradient Magnitude-based Token Selection Improves RLVR Training for LLM Reasoning
Reinforcement learning (RL), particularly RL with Verifiable Rewards (RLVR), has recently emerged as a central paradigm for enhancing large language models' (LLMs) reasoning abilities, demonstrating remarkable effectiveness across reasoning tasks. Recent studies suggest that high-entropy tokens play an exceptionally important role in model training, since training with only the highest 20%…
We haven't written up this one. arXiv cs.AI has the full story — the link below goes straight to it.
