AI Paper Atlas

Research map / Reinforcement learning / Policy gradient and actor-critic

GEPO: Group Expectation Policy Optimization for Stable Heterogeneous Reinforcement Learning

Han Zhang, RuibinZheng, ZEXUAN YI, Zhuo Zhang, Hanyang Peng, Hui Wang, Jiayin Qi, Binxing Fang, Ruifeng Xu, Yue Yu

ICLR 2026 · Reinforcement learning › Policy gradient and actor-critic · approach: Model-based RL with guarantees · 6 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use group-level importance weights (GEPO) instead of token-level (GRPO)GRPO · Online RL, Qwen3-8B, 4k, averageavg Pass@1: 61.8 -> 65.0Improved
use group-level importance weights (GEPO) instead of token-level (GRPO)GRPO · Online RL, Qwen3-1.7B, 4k, averageavg Pass@1: 49.9 -> 51.1Improved
use group-level instead of sequence-level importance weightsGSPO · Hetero RL max delay 64, Qwen3-8B, last checkpointavg Pass@1: 46.4 -> 60.8Improved
use group-level instead of sequence-level importance weightsGSPO · Hetero RL max delay 64, Qwen3-8B, best checkpointavg Pass@1: 58.4 -> 62.6Improved
use group-level instead of token-level importance weightsGRPO · Hetero RL max delay 64, Qwen3-8B, last checkpointavg Pass@1: 55.8 -> 60.8Improved
use sequence-level instead of token-level weightsGRPO · Hetero RL max delay 64, last checkpointavg Pass@1: 55.8 -> 46.4Hurt
use sequence-level instead of token-level weightsGRPO · Online RL, Qwen3-8B, 4k, averageavg Pass@1: 61.8 -> 60.5No clear effect
use BNPO or Dr.GRPO instead of GRPOGRPO · Hetero RL max delay 64, Qwen3-8Bavg Pass@1: 56.6 -> 55.9 (BNPO), 56.0 (Dr.GRPO)No clear effect
increase network latency (max delay 8 to 64 steps) (ablation)GEPO · Hetero RL training dynamicsKL divergence / eval accuracyHurt
higher latency raises KL divergence and importance weight varianceimportance sampling · Hetero RL, correlation analysiscorrelation: r = 0.96 (IW variance vs KL)Hurt
use group-expectation weight instead of 1/qimportance weights · Bias analysismean bias: 0.075 -> 0.072No clear effect
set KL coefficient to 0.005 and group size 8 (ablation)GEPO · Hetero RL hyperparameter sweep, Qwen3-1.7B, 2kavg best/lastImproved

Later papers that compared with this method

1 later accepted paper reports GEPO as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy