AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

When Preference Labels Fall Short: Aligning Diffusion Models from Real Data

Weiyan Chen, Weijian Deng, Yao Xiao, Weijie Tu, ZiYi Dong, Ibrahim Radwan, Liang Lin, Pengxu Wei

ICML 2026 · Image generation › Fast diffusion sampling · approach: Diffusion models as perception tools · 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
use 512 real-image preference pairs instead of 851k annotated pairs (Diffusion-DPO)SD-1.5 · Pick-a-Pic v2 testHPSv3: 6.80 -> 7.33Improved
use 512 real-image preference pairs instead of Diffusion-DPOSD-3.5-M · DrawBenchDeQA: 3.96 -> 4.26Improved
use real-data pairs vs Diffusion-DPOSD-3.5-M · DrawBenchUnifiedReward: 3.47 -> 3.46No clear effect
add stage 1 distribution alignment before DPO (ablation)two-stage method · Pick-a-Pic v2 testHPSv3: 6.11 -> 7.33Improved
add stage 2 DPO on contrastive pairs after stage 1 (ablation)two-stage method · Pick-a-Pic v2 testPickScore: 20.87 -> 21.04Improved
increase training pairs from 256 to 512 (ablation)two-stage method · Pick-a-Pic v2 testPickScore, HPSv3, DeQA, aestheticImproved
increase training pairs beyond 512 (to 1024) (ablation)two-stage method · Pick-a-Pic v2 testPickScore, HPSv3, DeQA, aestheticNo clear effect
apply real-data post-training on top of Diffusion-DPODiffusion-DPO (SD-1.5) · Pick-a-Pic v2 testall metricsImproved
apply real-data post-training on top of FlowGRPOFlowGRPO (SD-3.5-M) · DrawBenchall metricsImproved
use generated-image sources (Civitai-top) instead of real HPDv3 images (ablation)two-stage method · Pick-a-Pic v2 testHPSv3: 7.33 -> 8.67Improved
replace SD-1.5 inpainting with PixArt-alpha inpainting for negatives (ablation)two-stage method · Pick-a-Pic v2 testHPSv3: 7.33 -> 7.16No clear effect
replace inpainting with caption text-to-image negatives (ablation)two-stage method · Pick-a-Pic v2 testHPSv3: 7.33 -> 7.09No clear effect

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