AI Paper Atlas

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

Preference-Calibrated Optimization with Score-Level Distribution Alignment for Text-to-Image Diffusion Model Unlearning

Xiuyuan Wang, Weiming Liu, Hongyu Cai, Xin Gao, Fan Wang, Chaochao Chen, Xiaolin Zheng

ICML 2026 · Image generation › Text-to-image diffusion · approach: Inference-time T2I improvement

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 score-level distribution alignment (beta from 0 to 3e-6) (ablation)POSDA · UnlearnCanvas AbstractionismUA/IRA/CRA: beta 0 -> 3e-6Improved
increase alignment weight beyond optimum (3e-6 to 3e-5) (ablation)POSDA · UnlearnCanvas AbstractionismUA: 3e-6 -> 3e-5Hurt
add dynamic anchors (k = 3) to the naive reward (ablation)POSDA · UnlearnCanvas AbstractionismUA convergence and utilityImproved
use a single anchor (k = 1) instead of k = 3 (ablation)POSDA · UnlearnCanvas AbstractionismUA: k=1 vs k=3Hurt
use multiple anchors (k = 3) for faster erasure (ablation)POSDA · UnlearnCanvas Abstractionismtraining epochs to convergeImproved
POSDA vs best baseline SAeUron on style Van GoghPOSDA · UnlearnCanvas Van GoghAvg (%): 94.89 -> 97.77Improved
POSDA vs best baseline on object ArchitecturesPOSDA · UnlearnCanvas ArchitecturesAvg (%): 88.39 -> 98.89Improved
POSDA vs EDiff on NSFW nudityPOSDA · I2P NER-0.5NER (%): 5.1 -> 5.0Improved
POSDA vs original SD on exposed-part detectionsPOSDA · I2P total detectionscount: 443 -> 63Improved
POSDA vs original SD on CLIP scorePOSDA · MS-COCOCLIP score: 26.54 -> 26.73Improved
POSDA vs original SD on FIDPOSDA · MS-COCOFID: 52.00 -> 52.76No clear effect
ESD erasure vs POSDA on AbstractionismESD · UnlearnCanvas AbstractionismUA (%): 100.00 vs 98.33; Avg 91.43 vs 99.21Mixed

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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