{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/winoground-probing-vision-and-language-models","title":"Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality","arxiv_id":"2204.03162","date":"2022-04-07","proceeding":"CVPR 2022 1","authors":["Tristan Thrush","Ryan Jiang","Max Bartolo","Amanpreet Singh","Adina Williams","Douwe Kiela","Candace Ross"],"abstract":"We present a novel task and dataset for evaluating the ability of vision and language models to conduct visio-linguistic compositional reasoning, which we call Winoground. Given two images and two captions, the goal is to match them correctly - but crucially, both captions contain a completely identical set of words, only in a different order. The dataset was carefully hand-curated by expert annotators and is labeled with a rich set of fine-grained tags to assist in analyzing model performance. We probe a diverse range of state-of-the-art vision and language models and find that, surprisingly, none of them do much better than chance. Evidently, these models are not as skilled at visio-linguistic compositional reasoning as we might have hoped. We perform an extensive analysis to obtain insights into how future work might try to mitigate these models' shortcomings. We aim for Winoground to serve as a useful evaluation set for advancing the state of the art and driving further progress in the field. The dataset is available at https://huggingface.co/datasets/facebook/winoground.","url_abs":"https://arxiv.org/abs/2204.03162v2","url_pdf":"https://arxiv.org/pdf/2204.03162v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"winoground-probing-vision-and-language-models","repo_url":"https://github.com/wangt-cn/eqben","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"winoground-probing-vision-and-language-models","repo_url":"https://github.com/juletx/spatial-reasoning","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"visual-reasoning","task_name":"Visual Reasoning"}],"methods":[],"datasets_introduced":[{"slug":"winoground","name":"Winoground","full_name":""}],"methods_introduced":[],"results":[{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"UNITER large","rank_in_archive_order":43,"of":114,"metrics":{"Group Score":"10.50","Image Score":"14.00","Text Score":"38.00"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"VinVL","rank_in_archive_order":44,"of":114,"metrics":{"Group Score":"14.50","Image Score":"17.75","Text Score":"37.75"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"ViLLA large","rank_in_archive_order":45,"of":114,"metrics":{"Group Score":"11.00","Image Score":"13.25","Text Score":"37.00"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"ViLT (ViT-B/32)","rank_in_archive_order":53,"of":114,"metrics":{"Group Score":"9.25","Image Score":"14.00","Text Score":"34.75"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"FLAVA (ITM)","rank_in_archive_order":59,"of":114,"metrics":{"Group Score":"14.25","Image Score":"20.50","Text Score":"32.25"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"UNITER base","rank_in_archive_order":60,"of":114,"metrics":{"Group Score":"10.00","Image Score":"13.25","Text Score":"32.25"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"CLIP (ViT-B/32)","rank_in_archive_order":65,"of":114,"metrics":{"Group Score":"8.00","Image Score":"10.50","Text Score":"30.75"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"ViLLA base","rank_in_archive_order":69,"of":114,"metrics":{"Group Score":"8.00","Image Score":"12.00","Text Score":"30.00"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"FLAVA (contrastive)","rank_in_archive_order":85,"of":114,"metrics":{"Group Score":"9.00","Image Score":"13.50","Text Score":"25.25"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"Random chance","rank_in_archive_order":86,"of":114,"metrics":{"Group Score":"16.67","Image Score":"25.00","Text Score":"25.00"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"ViLBERT base","rank_in_archive_order":90,"of":114,"metrics":{"Group Score":"4.75","Image Score":"7.25","Text Score":"23.75"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"VSE++ (COCO, ResNet)","rank_in_archive_order":93,"of":114,"metrics":{"Group Score":"4.00","Image Score":"8.00","Text Score":"22.75"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"VSRN (Flickr30k)","rank_in_archive_order":98,"of":114,"metrics":{"Group Score":"3.50","Image Score":"5.00","Text Score":"20.00"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"VSE++ (Flickr30k, ResNet)","rank_in_archive_order":99,"of":114,"metrics":{"Group Score":"2.75","Image Score":"5.00","Text Score":"20.00"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"VSE++ (Flickr30k, VGG)","rank_in_archive_order":100,"of":114,"metrics":{"Group Score":"4.50","Image Score":"6.25","Text Score":"19.75"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"UniT (ITM finetuned)","rank_in_archive_order":101,"of":114,"metrics":{"Group Score":"4.00","Image Score":"6.25","Text Score":"19.50"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"LXMERT","rank_in_archive_order":102,"of":114,"metrics":{"Group Score":"4.00","Image Score":"7.00","Text Score":"19.25"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"VSE++ (COCO, VGG)","rank_in_archive_order":105,"of":114,"metrics":{"Group Score":"3.50","Image Score":"5.50","Text Score":"18.75"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"VSRN (COCO)","rank_in_archive_order":106,"of":114,"metrics":{"Group Score":"3.75","Image Score":"7.00","Text Score":"17.50"},"uses_additional_data":false},{"leaderboard":"/sota/visual-reasoning-on-winoground","task":"Visual Reasoning","dataset":"Winoground","model":"VisualBERT base","rank_in_archive_order":110,"of":114,"metrics":{"Group Score":"1.50","Image Score":"2.50","Text Score":"15.50"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=2204.03162","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2204.03162"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/wangt-cn/eqben","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/juletx/spatial-reasoning","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran":3,"unverified":1},"by_repo_kind":{"listed":{"samples":4,"ran":3,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"e21bb85b8e1b4836","entry":"do_nms","repo":"juletx/spatial-reasoning","repo_kind":"listed","path":"visual-spatial-reasoning/feature_extraction/lxmert/modeling_frcnn.py","file_url":"https://github.com/juletx/spatial-reasoning/blob/HEAD/visual-spatial-reasoning/feature_extraction/lxmert/modeling_frcnn.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"e21bb85b8e1b4836"}},{"code_sha256_prefix":"54a2fff601881b32","entry":"pad_list_tensors","repo":"juletx/spatial-reasoning","repo_kind":"listed","path":"visual-spatial-reasoning/feature_extraction/lxmert/modeling_frcnn.py","file_url":"https://github.com/juletx/spatial-reasoning/blob/HEAD/visual-spatial-reasoning/feature_extraction/lxmert/modeling_frcnn.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"54a2fff601881b32"}},{"code_sha256_prefix":"77586e7080d66922","entry":"tryload","repo":"juletx/spatial-reasoning","repo_kind":"listed","path":"visual-spatial-reasoning/feature_extraction/lxmert/extracting_data.py","file_url":"https://github.com/juletx/spatial-reasoning/blob/HEAD/visual-spatial-reasoning/feature_extraction/lxmert/extracting_data.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"77586e7080d66922"}},{"code_sha256_prefix":"ea0acbc5664a2925","entry":"norm_box","repo":"juletx/spatial-reasoning","repo_kind":"listed","path":"visual-spatial-reasoning/feature_extraction/lxmert/modeling_frcnn.py","file_url":"https://github.com/juletx/spatial-reasoning/blob/HEAD/visual-spatial-reasoning/feature_extraction/lxmert/modeling_frcnn.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"ea0acbc5664a2925"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}