AI Paper Atlas

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

Direct Diffusion Score Preference Optimization via Stepwise Contrastive Policy-Pair Supervision

Dohyun Kim, Seungwoo Lyu, Seung Kim, Paul Hongsuck Seo

ECCV 2026 · Image generation › Text-to-image diffusion · approach: Preference optimization for diffusion · 2 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 forward-posterior targets with contrastive-policy-pair targets in DSPODSPO · T2I alignment (SD-1.4)GenEval: .4841 -> .4978Improved
replace forward-process targets with DDSPO stepwise targets vs Diffusion DPODiffusion DPO · T2I alignment (SD-1.4)GenEval: .4841 -> .5045Improved
DDSPO versus D-DPO on image distribution shiftDiffusion DPO · MS-COCO 30K (SD-1.4)FID: 18.02 -> 16.39Improved
DSPO+CPP vs DDSPO on CompBenchDDSPO · T2I alignment (SD-1.4)T2I-CompBench: .3823 vs .3854Mixed
apply DDSPO to SDXLSDXL · T2I alignmentGenEval: .5229 -> .6049Improved
apply DDSPO to SD3-MSD3-M · T2I alignmentT2I-CompBench: .5155 -> .5740Improved
DDSPO without human annotation vs CaPOSD3-M · T2I alignmentGenEval: .7100 -> .7384Improved
train DDSPO with Pick-a-Pic (DD-CPP) vs D-DPOSDXL · Aesthetic qualityHPSv2: 28.55 -> 29.09Improved
use SPO instead of DDSPO (SDXL)SDXL · Aesthetic qualityPickScore: 22.71 -> 23.11Hurt
use random token removal instead of LLaMA (AQ) degraded prompts (ablation)DDSPO · T2I alignment (SD-1.4)GenEval: .4758 -> .5045Improved
use LLaMA (AQ) degraded prompts instead of random removal (ablation)DDSPO · Aesthetic quality (SD-1.4)HPSv2: 27.28 -> 27.51Improved
generate explicit negative images instead of reusing random positive image (ablation)DDSPO · T2I alignment (SD-1.4)GenEval: .4866 -> .5045Improved

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