Papers › Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality

Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality

7 Apr 2022CVPR 2022 1arXiv:2204.03162archive 2025-07-28

Tristan Thrush, Ryan Jiang, Max Bartolo, Amanpreet Singh, Adina Williams, Douwe Kiela, Candace Ross

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.

PaperPDFConference PDFCodeCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 4 code samples harvested from 1 repository linked to this paper: 4 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 0 violated, 3 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified. Where Syntology's instrument failed: 1 where our contract draft was wrong; failures of Syntology's instrument, not of the code.

By repository: community (archive-listed): 4 samples from 1 repository, 4 ran (3 with no instrument failure; 1 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

No sample is linked to this paper from 1 of its repositories that Syntology harvested for other papers:

Levels written on 2026-09-24 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

wangt-cn/eqben mentioned on GitHubpytorchApache-2.0Syntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to Equivariant Similarity for Vision-Language Foundation Models (arXiv:2303.14465; that paper's own run record: 1 ran (of which 1 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified; the one sample that ran constructed an object rather than computing a result); nothing checks that they implement this paper's method. report
juletx/spatial-reasoning pytorchMITSyntology: 4 samples, 4 ran (1 instrument-failed) · 0 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.

Code Syntology ran Syntology

4 samples harvested: 4 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 0 violated, 3 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

1ran · our draft was wrong
3ran

Licence: 3 of the 4 samples are pointer only, meaning Syntology does not serve that copy's text. For 3 of them, this paper's copy is under a permissive licence and cleared, but the stored copy of the same code that Syntology serves is not; the licence label's tooltip gives Syntology's recorded reason. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from juletx/spatial-reasoning. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: 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. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

juletx/spatial-reasoning community (archive-listed) 4 samples, 4 ran (1 instrument-failed) · 0 unverified
do_nms in 12 papers visual-spatial-reasoning/feature_extraction/lxmert/modeling_frcnn.py ran MIT (permissive) here; the stored copy is pointer only · e21bb85b8e1b4836 · report fixture: signature boxes:Tensor/2/float, scores:Tensor/2/float, image_shape:tuple/1/int, score_thresh:float/0/float, nms_thresh:float/0/float, mind:int/0/int, maxd:int/0/int · returned tuple · level recorded 2026-09-24
norm_box in 12 papers visual-spatial-reasoning/feature_extraction/lxmert/modeling_frcnn.py ran fingerprinted MIT (permissive) · ea0acbc5664a2925 · report fixture: signature boxes:Tensor/3/float, raw_sizes:Tensor/2/float · returned Tensor · level recorded 2026-09-24
pad_list_tensors in 12 papers visual-spatial-reasoning/feature_extraction/lxmert/modeling_frcnn.py ran MIT (permissive) here; the stored copy is pointer only · 54a2fff601881b32 · report fixture: signature list_tensors:list/1/Tensor, preds_per_image:list/1/int, max_detections:int/0/int, return_tensors:str/0/str, padding:str/0/str, pad_value:int/0/int, location:str/0/str · returned Tensor · level recorded 2026-09-24
tryload in 6 papers visual-spatial-reasoning/feature_extraction/lxmert/extracting_data.py ran · our draft was wrong MIT (permissive) here; the stored copy is pointer only · 77586e7080d66922 · report fixture: signature stream:StringIO/0/? · returned list · level recorded 2026-09-24

Tasks

Visual Reasoning

Datasets

Introduced by this paper, per the archive.

Winoground

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Visual Reasoning Winoground UNITER large Group Score 10.50 #43 of 114 Archive leaderboard report
Visual Reasoning Winoground UNITER large Image Score 14.00 #43 of 114 Archive leaderboard report
Visual Reasoning Winoground UNITER large Text Score 38.00 #43 of 114 Archive leaderboard report
Visual Reasoning Winoground VinVL Group Score 14.50 #44 of 114 Archive leaderboard report
Visual Reasoning Winoground VinVL Image Score 17.75 #44 of 114 Archive leaderboard report
Visual Reasoning Winoground VinVL Text Score 37.75 #44 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLLA large Group Score 11.00 #45 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLLA large Image Score 13.25 #45 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLLA large Text Score 37.00 #45 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLT (ViT-B/32) Group Score 9.25 #53 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLT (ViT-B/32) Image Score 14.00 #53 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLT (ViT-B/32) Text Score 34.75 #53 of 114 Archive leaderboard report
Visual Reasoning Winoground FLAVA (ITM) Group Score 14.25 #59 of 114 Archive leaderboard report
Visual Reasoning Winoground FLAVA (ITM) Image Score 20.50 #59 of 114 Archive leaderboard report
Visual Reasoning Winoground FLAVA (ITM) Text Score 32.25 #59 of 114 Archive leaderboard report
Visual Reasoning Winoground UNITER base Group Score 10.00 #60 of 114 Archive leaderboard report
Visual Reasoning Winoground UNITER base Image Score 13.25 #60 of 114 Archive leaderboard report
Visual Reasoning Winoground UNITER base Text Score 32.25 #60 of 114 Archive leaderboard report
Visual Reasoning Winoground CLIP (ViT-B/32) Group Score 8.00 #65 of 114 Archive leaderboard report
Visual Reasoning Winoground CLIP (ViT-B/32) Image Score 10.50 #65 of 114 Archive leaderboard report
Visual Reasoning Winoground CLIP (ViT-B/32) Text Score 30.75 #65 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLLA base Group Score 8.00 #69 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLLA base Image Score 12.00 #69 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLLA base Text Score 30.00 #69 of 114 Archive leaderboard report
Visual Reasoning Winoground FLAVA (contrastive) Group Score 9.00 #85 of 114 Archive leaderboard report
Visual Reasoning Winoground FLAVA (contrastive) Image Score 13.50 #85 of 114 Archive leaderboard report
Visual Reasoning Winoground FLAVA (contrastive) Text Score 25.25 #85 of 114 Archive leaderboard report
Visual Reasoning Winoground Random chance Group Score 16.67 #86 of 114 Archive leaderboard report
Visual Reasoning Winoground Random chance Image Score 25.00 #86 of 114 Archive leaderboard report
Visual Reasoning Winoground Random chance Text Score 25.00 #86 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLBERT base Group Score 4.75 #90 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLBERT base Image Score 7.25 #90 of 114 Archive leaderboard report
Visual Reasoning Winoground ViLBERT base Text Score 23.75 #90 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (COCO, ResNet) Group Score 4.00 #93 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (COCO, ResNet) Image Score 8.00 #93 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (COCO, ResNet) Text Score 22.75 #93 of 114 Archive leaderboard report
Visual Reasoning Winoground VSRN (Flickr30k) Group Score 3.50 #98 of 114 Archive leaderboard report
Visual Reasoning Winoground VSRN (Flickr30k) Image Score 5.00 #98 of 114 Archive leaderboard report
Visual Reasoning Winoground VSRN (Flickr30k) Text Score 20.00 #98 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (Flickr30k, ResNet) Group Score 2.75 #99 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (Flickr30k, ResNet) Image Score 5.00 #99 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (Flickr30k, ResNet) Text Score 20.00 #99 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (Flickr30k, VGG) Group Score 4.50 #100 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (Flickr30k, VGG) Image Score 6.25 #100 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (Flickr30k, VGG) Text Score 19.75 #100 of 114 Archive leaderboard report
Visual Reasoning Winoground UniT (ITM finetuned) Group Score 4.00 #101 of 114 Archive leaderboard report
Visual Reasoning Winoground UniT (ITM finetuned) Image Score 6.25 #101 of 114 Archive leaderboard report
Visual Reasoning Winoground UniT (ITM finetuned) Text Score 19.50 #101 of 114 Archive leaderboard report
Visual Reasoning Winoground LXMERT Group Score 4.00 #102 of 114 Archive leaderboard report
Visual Reasoning Winoground LXMERT Image Score 7.00 #102 of 114 Archive leaderboard report
Visual Reasoning Winoground LXMERT Text Score 19.25 #102 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (COCO, VGG) Group Score 3.50 #105 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (COCO, VGG) Image Score 5.50 #105 of 114 Archive leaderboard report
Visual Reasoning Winoground VSE++ (COCO, VGG) Text Score 18.75 #105 of 114 Archive leaderboard report
Visual Reasoning Winoground VSRN (COCO) Group Score 3.75 #106 of 114 Archive leaderboard report
Visual Reasoning Winoground VSRN (COCO) Image Score 7.00 #106 of 114 Archive leaderboard report
Visual Reasoning Winoground VSRN (COCO) Text Score 17.50 #106 of 114 Archive leaderboard report
Visual Reasoning Winoground VisualBERT base Group Score 1.50 #110 of 114 Archive leaderboard report
Visual Reasoning Winoground VisualBERT base Image Score 2.50 #110 of 114 Archive leaderboard report
Visual Reasoning Winoground VisualBERT base Text Score 15.50 #110 of 114 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections