English

EP-GRPO: Entropy-Progress Aligned Group Relative Policy Optimization with Implicit Process Guidance

Machine Learning 2026-05-07 v1 Artificial Intelligence

Abstract

Reinforcement learning with verifiable rewards (RLVR), particularly Group Relative Policy Optimization (GRPO), has advanced LLM reasoning. However, GRPO suffers from three credit assignment failures: uniform token-level granularity that ignores heterogeneous informational value, uniform polarity that penalizes correct steps and rewards incorrect ones, and zero-variance collapse that erases outcome-driven gradients. We systematically quantify these failures, revealing highly non-uniform token informativeness, widespread step-level polarity misalignment, and substantial training waste. To address these limitations, we propose Entropy-Progress Aligned GRPO (EP-GRPO), a framework that mines the model's intrinsic information flow for dense, self-supervised guidance. EP-GRPO integrates entropy-gated modulation to prioritize high entropy decision pivots, implicit process signals from policy divergence anchored to outcome advantages for directional token-level feedback without external reward models, and cumulative entropy mapping that enables progress-aligned advantage normalization, naturally maintaining gradient flow under zero reward variance. Extensive experiments on mathematical reasoning benchmarks demonstrate that EP-GRPO achieves superior accuracy and efficiency compared to GRPO and its variants. The code will be available.

Keywords

Cite

@article{arxiv.2605.04960,
  title  = {EP-GRPO: Entropy-Progress Aligned Group Relative Policy Optimization with Implicit Process Guidance},
  author = {Song Yu and Li Li and Wenwen Zhao and Zhisheng Yang},
  journal= {arXiv preprint arXiv:2605.04960},
  year   = {2026}
}

Comments

15 pages, 6 figures

R2 v1 2026-07-01T12:52:53.455Z