Hugging Face Trending Papers

Max Out GRPO Signal: Adaptive Trace Prefix Control for Hard Reasoning Problems

Read the original on Hugging Face Trending Papers →

Group Relative Policy Optimization (GRPO) stalls on a model's hardest problems: when no rollout in a group succeeds, the group-relative advantages vanish and the problem contributes no gradient, wasting the frontier examples we most want to learn from. Prepending a correct prefix of a reference solution raises the success rate, making prefix length a continuous knob on difficulty.

Summary generated by The Flow from the publisher's feed. The full article lives at Hugging Face Trending Papers.

arXiv AI
Jun 16

A First-Principles Derivation of LLM Policy Optimization: From Expected Reward to GRPO and Its Structural Extensions

arXiv:2606. 16733v1 Announce Type: new Abstract: Policy gradient algorithms for language models optimize the same objective $J(\theta) = \mathbb{E}*{\tau \sim p*\theta(\tau)}[R(\tau)]$, which has exactly two factors: the trajectory probability $p_\theta(\tau)$ and the reward $R(\tau)$.

By Jianghan Shen, Siqi Luo, Yue Li, Jiyao Liu, Wanying Qu, Yi Zhang, Ziyan Huang, Tianbin Li, Ming Hu, Xiaohong Liu, Yirong Chen, Junjun He