AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Expand and Prune: Maximizing Trajectory Diversity for Effective GRPO in Generative Models

Shiran Ge, Chenyi Huang, Yuang Ai, Qihang Fan, Huaibo Huang, Ran He

CVPR 2026 · Reinforcement learning › RL from feedback · approach: GRPO stability and variants · 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
Replace Flow-GRPO with Pro-GRPO (expand-and-prune)SD 3.5-M · DrawBenchIn-domain PickScore: 23.322 -> 24.008Improved
Replace Flow-GRPO with Pro-GRPOSD 3.5-M · HPSv2ImageReward (OOD): 1.463 -> 1.509Improved
Replace Flow-GRPO with Pro-GRPOSD 3.5-M · GenEval OverallOverall: 0.719 -> 0.726Improved
Replace Flow-GRPO with Pro-GRPOSD 3.5-M · GenEval Positionaccuracy: 0.288 -> 0.323Improved
Use Pro-GRPO-Flash (24->12 pruning, no expansion)SD 3.5-M · DrawBenchIn-domain PickScore and speedup: 23.322 -> 23.722; 1.41x speedupImproved
Replace DanceGRPO with Pro-GRPO (HPS reward)SD v1.4 · HPSv2.1 in-domainHPSv2.1: 0.369 -> 0.391Improved
Replace DanceGRPO with Pro-GRPO (HPS reward)SD v1.4 · HPSv2ImageReward (OOD): 1.079 -> 1.114Improved
Replace DanceGRPO with Pro-GRPO (HPS+CLIP reward)SD v1.4 · DrawBenchImageReward (OOD): 0.835 -> 0.974Improved
Pro-GRPO vs DanceGRPO (HPS reward) on PickScoreSD v1.4 · DrawBenchPickScore (OOD): 21.903 -> 21.896No clear effect
Increase initial group size Gmax from 32 to 48 (K=8) (ablation)Pro-GRPO SD v1.4 · HPSv2.1 rewardin-domain HPSv2.1: 0.386 -> 0.391Improved
Increase Gmax from 48 to 64 (K=8) (ablation)Pro-GRPO SD v1.4 · HPSv2.1 rewardin-domain HPSv2.1: 0.391 -> 0.393No clear effect
Move pruning checkpoints from {10,20} to {30,40} (ablation)Pro-GRPO SD v1.4 (32->8) · HPSv2.1 rewardin-domain HPSv2.1: 0.373 -> 0.391Improved

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