AI Paper Atlas

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

A Closed-Form Solution for Debiasing Vision-Language Models with Utility Guarantees Across Modalities and Tasks

Tangzheng Lian, Guanyu Hu, Yijing Ren, Dimitrios Kollias, Oya Celiktutan

CVPR 2026 · Vision-language and multimodal › CLIP and zero-shot · approach: Self-distillation for vision-language pretraining · 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
use spherical mean of LLM-generated variants plus anchor as group prototype instead of anchor only (ablation)Ours · Flickr30K MS@1000 (G)MS@1000: 13.4 -> 11.8Improved
use anchor plus variants prototype instead of mean of variants only (ablation)Ours · Flickr30K MS@1000 (G)MS@1000: 14.1 -> 11.8Improved
debias both image and text instead of text only (ablation)Ours · Flickr30K MS@1000 (G)MS@1000: 13.3 -> 11.8Improved
debias both image and text instead of image only (ablation)Ours · CelebA Delta Max EO (G x A)Delta Max EO: 41.7 -> 40.1Improved
swap LLM for prototype generation (DeepSeek v3.2, Gemini 2.5 Pro vs GPT-5) (ablation)Ours · Flickr30K MS@1000 / CelebA Delta Max EOfairness metrics: MS 12.0 and 11.8 vs 11.8No clear effect
project orthogonally to attribute subspace with Pareto-optimal closed-form (Ours) instead of Orth-ProjCLIP ViT-L/14 text-to-image retrieval · COCO2017 R@5R@5: 74.5 -> 81.1Improved
apply debiasing method to CLIPCLIP ViT-L/14 · Flickr30K MS@1000 (G)MS@1000: 20.3 -> 11.8Improved
apply debiasing method to CLIPCLIP ViT-L/14 · CelebA zero-shot F1F1: 54.0 -> 56.5Improved
apply debiasing method to CLIPCLIP ViT-L/14 · FACET Macro F1Macro F1: 70.8 -> 70.7No clear effect
apply debiasing to Stable Diffusion text embeddingsStable Diffusion v2.1 · gender-explicit prompts AccGAccG (%): 75.4 -> 74.6Mixed
debias only text embeddings (Orth-Proj) in SDStable Diffusion v2.1 · gender-explicit prompts AccGAccG (%): 75.4 -> 53.4Hurt

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