Papers › Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision

Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision

11 Feb 2021arXiv:2102.05918archive 2025-07-28

Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, YunHsuan Sung, Zhen Li, Tom Duerig

Pre-trained representations are becoming crucial for many NLP and perception tasks. While representation learning in NLP has transitioned to training on raw text without human annotations, visual and vision-language representations still rely heavily on curated training datasets that are expensive or require expert knowledge. For vision applications, representations are mostly learned using datasets with explicit class labels such as ImageNet or OpenImages. For vision-language, popular datasets like Conceptual Captions, MSCOCO, or CLIP all involve a non-trivial data collection (and cleaning) process. This costly curation process limits the size of datasets and hence hinders the scaling of trained models. In this paper, we leverage a noisy dataset of over one billion image alt-text pairs, obtained without expensive filtering or post-processing steps in the Conceptual Captions dataset. A simple dual-encoder architecture learns to align visual and language representations of the image and text pairs using a contrastive loss. We show that the scale of our corpus can make up for its noise and leads to state-of-the-art representations even with such a simple learning scheme. Our visual representation achieves strong performance when transferred to classification tasks such as ImageNet and VTAB. The aligned visual and language representations enables zero-shot image classification and also set new state-of-the-art results on Flickr30K and MSCOCO image-text retrieval benchmarks, even when compared with more sophisticated cross-attention models. The representations also enable cross-modality search with complex text and text + image queries.

PaperPDFCodeCode Syntology ranMethod as code

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 10 code samples harvested from 2 repositories linked to this paper: 8 ran (of which 6 constructed an object rather than computing a result; 7 with no instrument failure: 0 honoured, 0 violated, 7 with no contract checked; 1 where Syntology's instrument failed) · 2 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): 10 samples from 2 repositories, 8 ran (7 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 2026-09-22 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

MicPie/clasp mentioned on GitHubpytorchSyntology: 1 sample, 1 ran (0 instrument-failed) · 0 unverified report
facebookresearch/metaclip mentioned on GitHubpytorchSyntology: 9 samples, 7 ran (1 instrument-failed) · 2 unverified report
kakaobrain/coyo-dataset mentioned on GitHubpytorchSyntology: not harvested report
willard-yuan/video-text-retrieval-papers mentioned on GitHubSyntology: not harvested report
pwc-1/Paper-8 mindsporeSyntology: no sample linked to this paper (harvested for another paper). Syntology's graph links these samples to MathOdyssey: Benchmarking Mathematical Problem-Solving Skills in Large Language Models... (arXiv:2406.18321; that paper's own run record: 5 ran (of which 0 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified); nothing checks that they implement this paper's method. 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

10 samples harvested: 8 ran (of which 6 constructed an object rather than computing a result; 7 with no instrument failure: 0 honoured, 0 violated, 7 with no contract checked; 1 where Syntology's instrument failed) · 2 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; 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
7ran
2unverified

Licence: 9 of the 10 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from 2 repositories linked to this paper, official or community; the list below is grouped by repository, each group headed by its name and what it is to the paper. “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”. For 1 sample that ran, the graph records a signature this page does not show: for a class the fixture built without calling, the harness called the object with inputs it made up itself, so the recorded signature is the harness's probe, not the fixture's arguments. “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. A census tag appears when Syntology's census ran implementations with that name on one shared input per bucket: it gives how many were compared, from how many papers, and how many distinct outputs they produced, and names this sample's bucket when this sample was one of them.

MicPie/clasp community (archive-listed) 1 sample, 1 ran (0 instrument-failed) · 0 unverified
CLASP clasp/clasp.py ran · metamorphic tier: deterministic MIT (permissive) · badf4d34a0fb8563 · report fixture: returned CLASP · level recorded 2026-09-22
facebookresearch/metaclip community (archive-listed) 9 samples, 7 ran (1 instrument-failed) · 2 unverified
CLIPTextCfg src/mini_clip/model.py ran · metamorphic tier: deterministic licence not identified · pointer only · eb0a8c8d30313f06 · report fixture: returned CLIPTextCfg · level recorded 2026-09-22
CLIPVisionCfg in 5 papers src/mini_clip/model.py ran · metamorphic tier: deterministic licence not identified · pointer only · c3db51429b148b96 · report fixture: returned CLIPVisionCfg · level recorded 2026-09-22
LayerNorm in 121 papers
census: 5 implementations from 5 papers, 5 distinct outputs across 3 buckets, one shared input per bucket
src/mini_clip/model.py ran · metamorphic tier: invariant fingerprinted licence not identified · pointer only · 953eea434a052b02 · report fixture: signature x:Tensor/2/float · returned Tensor · level recorded 2026-09-22
ResidualAttentionBlock in 38 papers src/mini_clip/model.py ran · metamorphic tier: deterministic licence not identified · pointer only · 19aedb18f8c2e8ad · report fixture: returned ResidualAttentionBlock · level recorded 2026-09-22
Transformer in 173 papers src/mini_clip/model.py ran · metamorphic tier: deterministic licence not identified · pointer only · e93b074848e6cb6a · report fixture: returned Transformer · level recorded 2026-09-22
VisualTransformer in 12 papers src/mini_clip/model.py ran · metamorphic tier: deterministic licence not identified · pointer only · 29c96c042002f376 · report fixture: returned VisualTransformer · level recorded 2026-09-22
freeze_batch_norm_2d in 15 papers src/mini_clip/model.py ran · our draft was wrong licence not identified · pointer only · c9e7515a7f8b606b · report fixture: signature module:callable/0/?, module_match:dict/0/?, name:str/0/str · returned TestModule · level recorded 2026-09-22
CLIP in 25 papers src/mini_clip/model.py unverified licence not identified · pointer only · 349072ec3924682a · report
ModifiedResNet in 15 papers src/mini_clip/model.py unverified licence not identified · pointer only · 6da993cf6b63ba3a · report

Method as code Syntology

What Syntology's graph records as this paper's claimed method, lifted into code: each assembly names its bricks (functions and classes), each with its own role and its own verification level. An assembly is not a check that the code implements the method correctly or reproduces the paper's results. Levels are each brick's own verification level in Syntology's graph (level 0: no recorded run), listed brick by brick and never combined into one level; “ran” means the brick executed on a synthesized input, not that its output is correct, and the instrument figure counts failures of Syntology's instrument, not of the code. Where a model read the code against the paper's claim, each assembly's label verdict is one model's judgement, kept in this page's JSON twin; it is not calibrated, so no card prints it.

1 assembly on this paper: 1 harvested, 0 generated.

harvested Assembly for ALIGN (graph method: ALIGN). 1 brick; individually on a fixture: 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. The assembly has not been run end-to-end.

  1. CLASP entry · level 1 · ran · badf4d34a0fb8563

Each brick ends with its code_sha256, Syntology's identity for that exact code; click one to copy the get_code call that returns it. For agents, on Syntology's MCP service: assembly(arxiv_id="2102.05918") returns these assemblies with each brick's role and wiring (how to connect).

Tasks

Cross-Modal RetrievalFine-Grained Image ClassificationImage ClassificationImage-text RetrievalRepresentation LearningRetrievalText RetrievalZero-Shot Cross-Modal RetrievalZero-Shot Image ClassificationZero-Shot Transfer Image Classificationimage-classification

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Cross-Modal Retrieval COCO 2014 ALIGN Image-to-text R@1 77 #16 of 36 Archive leaderboard report
Cross-Modal Retrieval COCO 2014 ALIGN Image-to-text R@10 96.9 #16 of 36 Archive leaderboard report
Cross-Modal Retrieval COCO 2014 ALIGN Image-to-text R@5 93.5 #16 of 36 Archive leaderboard report
Cross-Modal Retrieval COCO 2014 ALIGN Text-to-image R@1 59.9 #16 of 36 Archive leaderboard report
Cross-Modal Retrieval COCO 2014 ALIGN Text-to-image R@10 89.8 #16 of 36 Archive leaderboard report
Cross-Modal Retrieval COCO 2014 ALIGN Text-to-image R@5 83.3 #16 of 36 Archive leaderboard report
Cross-Modal Retrieval Flickr30k ALIGN Image-to-text R@1 95.3 #9 of 27 Archive leaderboard report
Cross-Modal Retrieval Flickr30k ALIGN Image-to-text R@10 100 #9 of 27 Archive leaderboard report
Cross-Modal Retrieval Flickr30k ALIGN Image-to-text R@5 99.8 #9 of 27 Archive leaderboard report
Cross-Modal Retrieval Flickr30k ALIGN Text-to-image R@1 84.9 #9 of 27 Archive leaderboard report
Cross-Modal Retrieval Flickr30k ALIGN Text-to-image R@10 98.6 #9 of 27 Archive leaderboard report
Cross-Modal Retrieval Flickr30k ALIGN Text-to-image R@5 97.4 #9 of 27 Archive leaderboard report
Fine-Grained Image Classification Food-101 ALIGN Accuracy 95.88 #3 of 15 Archive leaderboard report
Fine-Grained Image Classification Oxford-IIIT Pet Dataset ALIGN Accuracy 96.19% #5 of 15 Archive leaderboard report
Fine-Grained Image Classification Stanford Cars ALIGN Accuracy 96.13% #6 of 83 Archive leaderboard report
Image Classification Flowers-102 ALIGN Accuracy 99.65% #7 of 52 Archive leaderboard report
Image Classification ImageNet ALIGN (EfficientNet-L2) Number of params 480M #34 of 1060 Archive leaderboard report
Image Classification ImageNet ALIGN (EfficientNet-L2) Top 1 Accuracy 88.64% #34 of 1060 Archive leaderboard report
Image Classification VTAB-1k ALIGN (50 hypers/task) Top-1 Accuracy 79.99 #1 of 34 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval COCO 2014 ALIGN Image-to-text R@1 58.6 #14 of 18 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval COCO 2014 ALIGN Image-to-text R@10 89.7 #14 of 18 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval COCO 2014 ALIGN Image-to-text R@5 83.0 #14 of 18 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval COCO 2014 ALIGN Text-to-image R@1 45.6 #14 of 18 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval COCO 2014 ALIGN Text-to-image R@10 78.6 #14 of 18 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval COCO 2014 ALIGN Text-to-image R@5 69.8 #14 of 18 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval Flickr30k ALIGN Image-to-text R@1 88.6 #14 of 22 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval Flickr30k ALIGN Image-to-text R@10 99.7 #14 of 22 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval Flickr30k ALIGN Image-to-text R@5 98.7 #14 of 22 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval Flickr30k ALIGN Text-to-image R@1 75.7 #14 of 22 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval Flickr30k ALIGN Text-to-image R@10 96.8 #14 of 22 Archive leaderboard report
Zero-Shot Cross-Modal Retrieval Flickr30k ALIGN Text-to-image R@5 93.8 #14 of 22 Archive leaderboard report
Zero-Shot Transfer Image Classification ImageNet ALIGN Accuracy (Private) 76.4 #16 of 23 Archive leaderboard report
Zero-Shot Transfer Image Classification ImageNet ALIGN Accuracy (Public) - #16 of 23 Archive leaderboard report
Zero-Shot Transfer Image Classification ImageNet V2 ALIGN Accuracy (Private) 70.1 #10 of 13 Archive leaderboard report
Zero-Shot Transfer Image Classification ImageNet V2 ALIGN Accuracy (Public) - #10 of 13 Archive leaderboard report
Zero-Shot Transfer Image Classification ImageNet-A ALIGN Accuracy (Private) 75.8 #11 of 13 Archive leaderboard report
Zero-Shot Transfer Image Classification ImageNet-A ALIGN Accuracy (Public) - #11 of 13 Archive leaderboard report
Zero-Shot Transfer Image Classification ImageNet-R ALIGN Accuracy 92.2 #9 of 12 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.

Methods

Introduced by this paper: ALIGN

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