AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Preference-Enhanced Reinforcement Learning for Pluralistic Image Inpainting

Peng Zhou, Muqi Huang, Tianshuo Qu, Jingyang Wang, kun Zhou, Chuan Li, Feng Shi, Shi Chen, Yun Xiong

ICML 2026 · Image generation › Text-to-image diffusion · approach: Preference optimization for diffusion

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
train Flux.1-fill with GRPO and preference reward instead of baseFlux.1-fill · FluxBenchImageReward x10^2: -32.53 -> 31.41Improved
train Flux.1-fill with GRPO and preference reward instead of baseFlux.1-fill · EditBenchImageReward x10^2: -33.36 -> 8.98Improved
train Flux.1-fill with GRPO and preference reward instead of baseFlux.1-fill · BrushBenchImageReward x10^2: 48.64 -> 55.03Improved
train SD1.5-inpainting with GRPO instead of baseSD1.5-inpainting · BrushBenchImageReward x10^2: 38.40 -> 53.86Improved
replace SFT with RL (GRPO) trainingFlux.1-fill · FluxBenchImageReward x10^2: -34.57 -> 31.41Improved
replace SFT with RL (GRPO) trainingSD1.5-inpainting · BrushBenchImageReward x10^2: 38.81 -> 53.86Improved
RL without ground-truth reconstruction constraintFlux.1-fill · FluxBenchPSNR (mask preservation): 6.993 -> 6.278Hurt
add boundary reward (BGCS) with fixed weights (ablation)Flux.1-fill RL · FluxBenchBGCS: -0.0304 -> -0.0183Improved
add dynamic mask-ratio weighting to the reward (ablation)Flux.1-fill RL · FluxBenchBGCS: -0.0183 -> -0.0163Improved
add dynamic mask-ratio weighting to the reward (ablation)Flux.1-fill RL · FluxBenchHPSv3: 4.193 -> 4.461Improved
add dynamic weighting and boundary reward (ablation)Flux.1-fill RL · BrushBenchBGCS: -0.1232 -> -0.1114Improved
RL with only quality and text rewards, no boundary term (ablation)Flux.1-fill RL · FluxBenchBGCS: -0.0369 -> -0.0304No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

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