AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Test-Time Debiasing with Probabilistic Prompts via Wasserstein Distance in Vision-Language Models

Chengye Wang, Yuyuan Li, XiaoHua Feng, Xiaolin Zheng, Chaochao Chen

ICML 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Test-time adaptation of VLMs

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 point-based debiasing with Wasserstein distributional alignment plus probabilistic promptsCLIP-ViT-B-P16 · UTKFace race, stereotype queriesMaxSkew: 1.0074 -> 0.3312Improved
replace point-based debiasing with Wasserstein distributional alignment plus probabilistic promptsCLIP-ViT-B-P16 · UTKFace gender, stereotype queriesMaxSkew: 0.8204 -> 0.0833Improved
compare against BendVLM on binary gender attributeCLIP-ViT-L-P14 · FairFace gender, stereotype queriesMaxSkew: 0.0952 (BendVLM) vs 0.1324 (W4D)Hurt
increase beta (weight of performance preservation loss) from 0 to 2.0 (ablation)W4D · CelebA/Facet zero-shot classificationWorst-Group AUC-ROC and MaxSkewMixed
increase gamma (orthogonality loss weight) from 0 to 3.0 (ablation)W4D · CelebA/Facet zero-shot classificationWorst-Group AUC-ROC and MaxSkewMixed
increase Monte Carlo sample size K from 16 to 256 (ablation)W4D probabilistic prompts · CelebA/Facet zero-shot classificationAUC and MaxSkewImproved
apply orthogonal projection or orthogonal calibration debiasingCLIP-ViT-B-P16 · FairFace gender, stereotype queriesMaxSkew: 0.6662 -> 1.3983 (Orth-Proj.)Hurt
use W4D in zero-shot classificationCLIP · CelebA, Facet zero-shot classificationWorst-Group AUC-ROCNo 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