AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Geometric-Mean Policy Optimization

Yuzhong Zhao, Yue Liu, Junpeng Liu, Jingye Chen, Xun Wu, Yaru Hao, Tengchao Lv, Shaohan Huang, Lei Cui, Qixiang Ye, Fang Wan, Furu Wei

ICLR 2026 · Reinforcement learning › RL from feedback · approach: GRPO stability and variants

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
replace arithmetic mean with geometric mean of token rewards (GMPO vs GRPO)GRPO, Qwen2.5-Math-7B · 5 math benchmarks avgPass@1 (%): 51.2 -> 52.7Improved
GMPO vs GRPODeepSeek-R1-Distill-Qwen-7B · 5 math benchmarks avgPass@1 (%): 59.3 -> 63.4Improved
GMPO vs GRPOQwen2.5-Math-1.5B · 5 math benchmarks avgPass@1 (%): 42.5 -> 43.9Improved
GMPO vs GRPOQwen2.5-VL-Instruct-7B · Geometry3KPass@1 (%): 53.3 -> 54.7Improved
GMPO vs GRPOQwen3-32B MoE · MATH500Pass@1 (%): 94.6 -> 96.7Improved
add 1/|o| length normalization to the geometric product (ablation)GMPO · 5 math benchmarks avgPass@1 (%): 52.0 -> 52.7Improved
add clipping to GMPO (ablation)GMPO · 5 math benchmarks avgPass@1 (%): 52.3 -> 52.7Improved
token-level clipping instead of sequence-level clipping (ablation)GMPO · 5 math benchmarks avgPass@1 (%): 52.6 (seq-clip) vs 52.7Mixed
widen clip range to (e^-0.4, e^0.4) from (e^-0.2, e^0.2) (ablation)GMPO · 5 math benchmarks avgPass@1 (%): 52.4 -> 52.7Improved
widen clip range further to (e^-0.8, e^0.8) (ablation)GMPO · 5 math benchmarks avgPass@1 (%): 52.7 -> 52.1Hurt

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