AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

GR-GRPO: Graph-Diffused Credit for Autoregressive Image RL Alignment

Zheyu Zhang, Peng-Tao Jiang, Tianyi Zheng, Jian Zhang, Jinwei Chen, Bo Li

ECCV 2026 · Reinforcement learning › RL from feedback · approach: GRPO stability and variants · 0 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
add graph-regularized credit (GR-GRPO) to GRPO (ablation)Janus-Pro-1B GRPO · GenEvaloverall success rate: 0.84 -> 0.86Improved
add difficulty-aware budget allocation to GRPO alone (ablation)Janus-Pro-1B GRPO · GenEvaloverall success rate: 0.84 -> 0.84No clear effect
add difficulty-aware budget allocation on top of graph regularization (ablation)GR-GRPO · GenEvaloverall success rate: 0.86 -> 0.87Improved
use confidence-adaptive gate instead of constant gate (ablation)GR-GRPO · GenEvaloverall success rate: 0.83 -> 0.87Improved
propagate negative evidence in addition to positive (ablation)GR-GRPO · GenEvaloverall success rate: 0.87 -> 0.84Hurt
use K=64 instead of K=32 or 128 (ablation)GR-GRPO · GenEvaloverall success rate: 0.84 (K=32), 0.86 (K=128) -> 0.87 (K=64)Improved
use intermediate diffusion strength 0.5 instead of 0.3 or 0.7 (ablation)GR-GRPO · GenEvaloverall success rate: 0.84 (0.3), 0.84 (0.7) -> 0.87 (0.5)Improved
GR-GRPO vs vanilla GRPO on Janus-Pro-1BJanus-Pro-1B · GenEvaloverall success rate: 0.84 -> 0.87Improved
GR-GRPO vs vanilla GRPO on Janus-Pro-7BJanus-Pro-7B · GenEvaloverall success rate: 0.87 -> 0.92Improved
GR-GRPO vs STAGE on Color Attr.Janus-Pro-7B · GenEval (Color Attr.)success rate: 0.79 -> 0.74Hurt
GR-GRPO vs STAGE on DrawBench UnifiedRewardJanus-Pro-7B · DrawBenchUnifiedReward: 3.10 -> 3.07Hurt
GR-GRPO vs vanilla GRPO on DrawBenchJanus-Pro-1B · DrawBenchHPSv2.1: 27.22 -> 27.55Improved

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