AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Bias in Gender Bias Benchmarks: How Spurious Features Distort Evaluation

Yusuke Hirota, Ryo Hachiuma, Boyi Li, Ximing Lu, Michael Ross Boone, Boris Ivanovic, Yejin Choi, Marco Pavone, Yu-Chiang Frank Wang, Noa Garcia, Yuta Nakashima, Chao-Han Huck Yang

ICCV 2025 · Vision-language and multimodal › Human and embodied modeling · approach: Human-like visual cognition in VLMs · 1 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
perturb background by blurring (weak)generative VLMs bias evaluation · COCO-gender YGaprelative difference Delta: 165.84% for LLaVA-OneVision-7BHurt
mask 10% of objectsgenerative VLMs bias evaluation · FACET / MIAP YGaprelative difference Delta: 40.98% (Qwen2-VL-7B, FACET), 48.86% (InternVL-2.5-8B, MIAP)Hurt
perturb lighting (brightness)bias evaluation · all four benchmarksrelative difference Delta: mostly below 10% for generative VLMs, below 2% for CLIPNo clear effect
shift hue (color perturbation)bias evaluation · MIAP YGaprelative difference Delta: 20.88% (EAGLE-8B, strong)Mixed
perturb object and background features for CLIP variantsCLIP MaxSkew@1000 · FACET / PHASErelative difference Delta: 25.03% object (ViT-B/32 FACET), 51.01% background (ViT-B/32 PHASE)Hurt
stronger spurious correlation of a feature with genderbias evaluation · generative VLMs vs CLIP variantscorrelation of Acc_b with Delta: r = 0.31 (generative), r = 0.54 (CLIP)Improved
perturb object featuresmodel bias rankings · FACET YGap rankingbias rankingHurt
use ResNet-50 instead of ConvNeXt-B for feature gender classifier (ablation)spurious feature detection · four benchmarksgender prediction accuracy: COCO-gender object 76.3 -> 76.3No 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