{"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/scaling-up-visual-and-vision-language","title":"Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision","arxiv_id":"2102.05918","date":"2021-02-11","proceeding":null,"authors":["Chao Jia","Yinfei Yang","Ye Xia","Yi-Ting Chen","Zarana Parekh","Hieu Pham","Quoc V. Le","YunHsuan Sung","Zhen Li","Tom Duerig"],"abstract":"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.","url_abs":"https://arxiv.org/abs/2102.05918v2","url_pdf":"https://arxiv.org/pdf/2102.05918v2.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":"scaling-up-visual-and-vision-language","repo_url":"https://github.com/MicPie/clasp","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"scaling-up-visual-and-vision-language","repo_url":"https://github.com/facebookresearch/metaclip","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"scaling-up-visual-and-vision-language","repo_url":"https://github.com/kakaobrain/coyo-dataset","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"scaling-up-visual-and-vision-language","repo_url":"https://github.com/willard-yuan/video-text-retrieval-papers","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"scaling-up-visual-and-vision-language","repo_url":"https://github.com/pwc-1/Paper-8/tree/main/align","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null}],"tasks":[{"task_slug":"cross-modal-retrieval","task_name":"Cross-Modal Retrieval"},{"task_slug":"fine-grained-image-classification","task_name":"Fine-Grained Image Classification"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"image-text-retrieval","task_name":"Image-text Retrieval"},{"task_slug":"representation-learning","task_name":"Representation Learning"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"text-retrieval","task_name":"Text Retrieval"},{"task_slug":"zero-shot-cross-modal-retrieval","task_name":"Zero-Shot Cross-Modal Retrieval"},{"task_slug":"zero-shot-image-classification","task_name":"Zero-Shot Image Classification"},{"task_slug":"zero-shot-transfer-image-classification","task_name":"Zero-Shot Transfer Image Classification"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[{"slug":"align","name":"ALIGN","full_name":"ALIGN"}],"results":[{"leaderboard":"/sota/cross-modal-retrieval-on-coco-2014","task":"Cross-Modal Retrieval","dataset":"COCO 2014","model":"ALIGN","rank_in_archive_order":16,"of":36,"metrics":{"Image-to-text R@1":"77","Image-to-text R@10":"96.9","Image-to-text R@5":"93.5","Text-to-image R@1":"59.9","Text-to-image R@10":"89.8","Text-to-image R@5":"83.3"},"uses_additional_data":true},{"leaderboard":"/sota/cross-modal-retrieval-on-flickr30k","task":"Cross-Modal Retrieval","dataset":"Flickr30k","model":"ALIGN","rank_in_archive_order":9,"of":27,"metrics":{"Image-to-text R@1":"95.3","Image-to-text R@10":"100","Image-to-text R@5":"99.8","Text-to-image R@1":"84.9","Text-to-image R@10":"98.6","Text-to-image R@5":"97.4"},"uses_additional_data":true},{"leaderboard":"/sota/fine-grained-image-classification-on-food-101","task":"Fine-Grained Image Classification","dataset":"Food-101","model":"ALIGN","rank_in_archive_order":3,"of":15,"metrics":{"Accuracy":"95.88"},"uses_additional_data":true},{"leaderboard":"/sota/fine-grained-image-classification-on-oxford-1","task":"Fine-Grained Image Classification","dataset":"Oxford-IIIT Pet Dataset","model":"ALIGN","rank_in_archive_order":5,"of":15,"metrics":{"Accuracy":"96.19%"},"uses_additional_data":true},{"leaderboard":"/sota/fine-grained-image-classification-on-stanford","task":"Fine-Grained Image Classification","dataset":"Stanford Cars","model":"ALIGN","rank_in_archive_order":6,"of":83,"metrics":{"Accuracy":"96.13%"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-flowers-102","task":"Image Classification","dataset":"Flowers-102","model":"ALIGN","rank_in_archive_order":7,"of":52,"metrics":{"Accuracy":"99.65%"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ALIGN (EfficientNet-L2)","rank_in_archive_order":34,"of":1060,"metrics":{"Number of params":"480M","Top 1 Accuracy":"88.64%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-vtab-1k-1","task":"Image Classification","dataset":"VTAB-1k","model":"ALIGN (50 hypers/task)","rank_in_archive_order":1,"of":34,"metrics":{"Top-1 Accuracy":"79.99"},"uses_additional_data":true},{"leaderboard":"/sota/zero-shot-cross-modal-retrieval-on-coco-2014","task":"Zero-Shot Cross-Modal Retrieval","dataset":"COCO 2014","model":"ALIGN","rank_in_archive_order":14,"of":18,"metrics":{"Image-to-text R@1":"58.6","Image-to-text R@10":"89.7","Image-to-text R@5":"83.0","Text-to-image R@1":"45.6","Text-to-image R@10":"78.6","Text-to-image R@5":"69.8"},"uses_additional_data":false},{"leaderboard":"/sota/zero-shot-cross-modal-retrieval-on-flickr30k","task":"Zero-Shot Cross-Modal Retrieval","dataset":"Flickr30k","model":"ALIGN","rank_in_archive_order":14,"of":22,"metrics":{"Image-to-text R@1":"88.6","Image-to-text R@10":"99.7","Image-to-text R@5":"98.7","Text-to-image R@1":"75.7","Text-to-image R@10":"96.8","Text-to-image R@5":"93.8"},"uses_additional_data":false},{"leaderboard":"/sota/zero-shot-transfer-image-classification-on-1","task":"Zero-Shot Transfer Image Classification","dataset":"ImageNet","model":"ALIGN","rank_in_archive_order":16,"of":23,"metrics":{"Accuracy (Private)":"76.4","Accuracy (Public)":"-"},"uses_additional_data":false},{"leaderboard":"/sota/zero-shot-transfer-image-classification-on-3","task":"Zero-Shot Transfer Image Classification","dataset":"ImageNet V2","model":"ALIGN","rank_in_archive_order":10,"of":13,"metrics":{"Accuracy (Private)":" 70.1","Accuracy (Public)":"-"},"uses_additional_data":false},{"leaderboard":"/sota/zero-shot-transfer-image-classification-on-5","task":"Zero-Shot Transfer Image Classification","dataset":"ImageNet-A","model":"ALIGN","rank_in_archive_order":11,"of":13,"metrics":{"Accuracy (Private)":"75.8","Accuracy (Public)":"-"},"uses_additional_data":false},{"leaderboard":"/sota/zero-shot-transfer-image-classification-on-4","task":"Zero-Shot Transfer Image Classification","dataset":"ImageNet-R","model":"ALIGN","rank_in_archive_order":9,"of":12,"metrics":{"Accuracy":"92.2"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2102.05918","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2102.05918"}},"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/facebookresearch/metaclip","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/willard-yuan/video-text-retrieval-papers","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/kakaobrain/coyo-dataset","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/pwc-1/Paper-8/tree/main/align","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/MicPie/clasp","reach":null}],"summary":{"ran":7,"ran_draft_wrong":1,"unverified":2},"by_repo_kind":{"listed":{"samples":10,"ran":8,"repositories":2}},"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":9,"samples":[{"code_sha256_prefix":"badf4d34a0fb8563","entry":"CLASP","repo":"MicPie/clasp","repo_kind":"listed","path":"clasp/clasp.py","file_url":"https://github.com/MicPie/clasp/blob/HEAD/clasp/clasp.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"badf4d34a0fb8563"}},{"code_sha256_prefix":"eb0a8c8d30313f06","entry":"CLIPTextCfg","repo":"facebookresearch/metaclip","repo_kind":"listed","path":"src/mini_clip/model.py","file_url":"https://github.com/facebookresearch/metaclip/blob/HEAD/src/mini_clip/model.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"eb0a8c8d30313f06"}},{"code_sha256_prefix":"c3db51429b148b96","entry":"CLIPVisionCfg","repo":"facebookresearch/metaclip","repo_kind":"listed","path":"src/mini_clip/model.py","file_url":"https://github.com/facebookresearch/metaclip/blob/HEAD/src/mini_clip/model.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"c3db51429b148b96"}},{"code_sha256_prefix":"953eea434a052b02","entry":"LayerNorm","repo":"facebookresearch/metaclip","repo_kind":"listed","path":"src/mini_clip/model.py","file_url":"https://github.com/facebookresearch/metaclip/blob/HEAD/src/mini_clip/model.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"953eea434a052b02"}},{"code_sha256_prefix":"19aedb18f8c2e8ad","entry":"ResidualAttentionBlock","repo":"facebookresearch/metaclip","repo_kind":"listed","path":"src/mini_clip/model.py","file_url":"https://github.com/facebookresearch/metaclip/blob/HEAD/src/mini_clip/model.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"19aedb18f8c2e8ad"}},{"code_sha256_prefix":"e93b074848e6cb6a","entry":"Transformer","repo":"facebookresearch/metaclip","repo_kind":"listed","path":"src/mini_clip/model.py","file_url":"https://github.com/facebookresearch/metaclip/blob/HEAD/src/mini_clip/model.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"e93b074848e6cb6a"}},{"code_sha256_prefix":"29c96c042002f376","entry":"VisualTransformer","repo":"facebookresearch/metaclip","repo_kind":"listed","path":"src/mini_clip/model.py","file_url":"https://github.com/facebookresearch/metaclip/blob/HEAD/src/mini_clip/model.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"29c96c042002f376"}},{"code_sha256_prefix":"c9e7515a7f8b606b","entry":"freeze_batch_norm_2d","repo":"facebookresearch/metaclip","repo_kind":"listed","path":"src/mini_clip/model.py","file_url":"https://github.com/facebookresearch/metaclip/blob/HEAD/src/mini_clip/model.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"c9e7515a7f8b606b"}},{"code_sha256_prefix":"349072ec3924682a","entry":"CLIP","repo":"facebookresearch/metaclip","repo_kind":"listed","path":"src/mini_clip/model.py","file_url":"https://github.com/facebookresearch/metaclip/blob/HEAD/src/mini_clip/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"349072ec3924682a"}},{"code_sha256_prefix":"6da993cf6b63ba3a","entry":"ModifiedResNet","repo":"facebookresearch/metaclip","repo_kind":"listed","path":"src/mini_clip/model.py","file_url":"https://github.com/facebookresearch/metaclip/blob/HEAD/src/mini_clip/model.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NOASSERTION","inline_ok":false,"mcp_get_code":{"code_sha256":"6da993cf6b63ba3a"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}