{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T10:24:19Z","timestamp":1777890259904,"version":"3.51.4"},"reference-count":71,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,10,19]]},"DOI":"10.1109\/iccv51701.2025.00808","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"8634-8644","source":"Crossref","is-referenced-by-count":0,"title":["Bias in Gender Bias Benchmarks: How Spurious Features Distort Evaluation"],"prefix":"10.1109","author":[{"given":"Yusuke","family":"Hirota","sequence":"first","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ryo","family":"Hachiuma","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Boyi","family":"Li","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ximing","family":"Lu","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Michael Ross","family":"Boone","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Boris","family":"Ivanovic","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yejin","family":"Choi","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Marco","family":"Pavone","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yu-Chiang Frank","family":"Wang","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Noa","family":"Garcia","sequence":"additional","affiliation":[{"name":"Osaka University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yuta","family":"Nakashima","sequence":"additional","affiliation":[{"name":"Osaka University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Chao-Han Huck","family":"Yang","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.emnlp-main.88"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.aacl-main.61"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1145\/3593013.3594095"},{"key":"ref4","first-page":"1","article-title":"Multimodal datasets: Misogyny, pornography, and malignant stereotypes","author":"Birhane","year":"2021","journal-title":"arXiv preprint"},{"key":"ref5","article-title":"Expanding performance boundaries of open-source multimodal models with model, data, and testtime scaling","author":"Chen","year":"2024","journal-title":"arXiv preprint"},{"key":"ref6","first-page":"2","article-title":"Dall-eval: Probing the reasoning skills and social biases of text-toimage generation models","author":"Cho","year":"2023","journal-title":"ICCV"},{"key":"ref7","article-title":"Debiasing visionlanguage models via biased prompts","author":"Chuang","year":"2023","journal-title":"arXiv preprint"},{"key":"ref8","article-title":"FairVLM: Mitigating bias in pre-trained vision-language models","author":"Dehdashtian","year":"2024","journal-title":"ICLR"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2021.3136898"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.eacl-long.41"},{"key":"ref11","first-page":"4","article-title":"Fair diffusion: Instructing text-to-image generation models on fairness","author":"Friedrich","year":"2023","journal-title":"arXiv preprint"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00672"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1145\/3292500.3330691"},{"key":"ref14","article-title":"Revealing and reducing gender biases in vision and language assistants (vlas)","author":"Girrbach","year":"2025","journal-title":"ICLR"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01863"},{"key":"ref16","first-page":"2","article-title":"Vision-language models performing zero-shot tasks exhibit gender-based disparities","volume-title":"ICCV Workshops","author":"Hall","year":"2023"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46493-0_38"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01309"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.986"},{"key":"ref20","first-page":"7","article-title":"Saner: Annotation-free societal attribute neutralizer for debiasing clip","author":"Hirota","year":"2025","journal-title":"ICLR"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01138"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.713"},{"key":"ref23","first-page":"2","article-title":"Openimages: A public dataset for large-scale multi-label and multi-class image classification","author":"Krasin","year":"2017","journal-title":"Dataset available from"},{"key":"ref24","article-title":"Llava-onevision: Easy visual task transfer","author":"Li","year":"2024","journal-title":"TMLR"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-10602-1_48"},{"key":"ref26","article-title":"Visual instruction tuning","author":"Liu","year":"2024","journal-title":"NeurIPS"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01167"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01029"},{"key":"ref29","first-page":"2","article-title":"Stable bias: Analyzing societal representations in diffusion models","author":"Luccioni","year":"2023","journal-title":"NeurIPS"},{"key":"ref30","first-page":"8","article-title":"Multimodal composite association score: Measuring gender bias in generative multimodal models","author":"Mandal","year":"2023","journal-title":"arXiv preprint"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00446"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1145\/3600211.3604711"},{"key":"ref33","first-page":"3","article-title":"Rectified linear units improve restricted boltzmann machines","author":"Nair","year":"2010","journal-title":"ICML"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.520"},{"key":"ref35","article-title":"Learning transferable visual models from natural language supervision","author":"Radford","year":"2021","journal-title":"ICML"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-emnlp.611"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01042"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.naacl-main.78"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.findings-acl.403"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-emnlp.66"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1145\/3461702.3462594"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.naacl-long.353"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00659"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P18-1238"},{"key":"ref45","first-page":"5","article-title":"Eagle: Exploring the design space for multimodal 11 ms with mixture of encoders","author":"Shi","year":"2025","journal-title":"ICLR"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01319"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.gebnlp-1.10"},{"key":"ref48","first-page":"8","article-title":"The biased artist: Exploiting cultural biases via homoglyphs in text-guided image generation models","author":"Struppek","year":"2022","journal-title":"arXiv preprint"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1109\/WACV57701.2024.00296"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.findings-acl.502"},{"key":"ref51","first-page":"2","article-title":"Directional bias amplification","author":"Wang","year":"2021","journal-title":"ICML"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00366"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.emnlp-main.151"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.findings-acl.160"},{"key":"ref55","article-title":"Qwen2-vl: Enhancing vision-language model\u2019s perception of the world at any resolution","author":"Wang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1109\/tpami.2026.3683747"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00541"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00308"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.213"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.878"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1145\/3593013.3594072"},{"key":"ref62","first-page":"3","article-title":"Genderbias-emph : Benchmarking gender bias in vision language models via counterfactual probing","author":"Xiao","year":"2024","journal-title":"arXiv preprint"},{"key":"ref63","first-page":"5","article-title":"mplugowl3: Towards long image-sequence understanding in multimodal large language models","author":"Ye","year":"2025","journal-title":"ICLR"},{"key":"ref64","first-page":"7","article-title":"Coca: Contrastive captioners are image-text foundation models","author":"Yu","year":"2022","journal-title":"TMLR"},{"key":"ref65","first-page":"2","article-title":"Understanding bias in large-scale visual datasets","author":"Zeng","year":"2025","journal-title":"NeurIPS"},{"key":"ref66","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01100"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52734.2025.00401"},{"key":"ref68","first-page":"8","article-title":"Auditing gender presentation differences in text-to-image models","author":"Zhang","year":"2023","journal-title":"arXiv preprint"},{"key":"ref69","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.01456"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D17-1323"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-20077-9_21"}],"event":{"name":"2025 IEEE\/CVF International Conference on Computer Vision (ICCV)","location":"Honolulu, HI, USA","start":{"date-parts":[[2025,10,19]]},"end":{"date-parts":[[2025,10,25]]}},"container-title":["2025 IEEE\/CVF International Conference on Computer Vision (ICCV)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11443115\/11443287\/11445936.pdf?arnumber=11445936","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T05:33:13Z","timestamp":1777613593000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11445936\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":71,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.00808","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}