{"url":"/sota/image-to-text-retrieval-on-coco","task":{"name":"Image-to-Text Retrieval","url":"/task/image-to-text-retrieval","note":null},"dataset":{"name":"COCO (Common Objects in Context)","url":"/dataset/coco"},"category":"Natural Language Processing","categories":["Natural Language Processing"],"category_note":null,"description":"**Image-text retrieval** is the process of retrieving relevant images based on textual descriptions or finding corresponding textual descriptions for a given image. This task is interdisciplinary, combining techniques from computer vision, and natural language processing. The primary challenge lies in bridging the *semantic gap* — the difference between how visual data is represented in images and how humans describe that information using language. To address this, many methods focus on learning a shared embedding space where both images and text can be represented in a comparable way, allowing their similarities to be measured and facilitating more accurate retrieval.\r\n\r\n**Source**: <span class=\"description-source\"> [Extending CLIP for Category-to-Image Retrieval in E-commerce](https://arxiv.org/abs/2112.11294)</span>","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["Recall@1","Recall@5","Recall@10"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Recall@1":"higher","Recall@5":"higher","Recall@10":"higher"}},"counts":{"rows":9,"rows_with_code":8,"rows_with_paper_page":9,"rows_dated":9,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"BLIP-2 (ViT-G, fine-tuned)","metrics":{"Recall@1":"85.4","Recall@10":"98.5","Recall@5":"97.0"},"uses_additional_data":false,"paper_date":"2023-01-30","paper":"/paper/blip-2-bootstrapping-language-image-pre","paper_url":"https://arxiv.org/abs/2301.12597v3","paper_title":"BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models","code":"https://github.com/huggingface/transformers","n_code_links":17,"syntology":{"n_ran":4,"n_unverified":4,"n_samples":8,"n_pointer_only_licence":1}},{"rank_in_archive_order":2,"model":"ONE-PEACE (ViT-G, w/o ranking)","metrics":{"Recall@1":"84.1","Recall@10":"98.3","Recall@5":"96.3"},"uses_additional_data":false,"paper_date":"2023-05-18","paper":"/paper/one-peace-exploring-one-general","paper_url":"https://arxiv.org/abs/2305.11172v1","paper_title":"ONE-PEACE: Exploring One General Representation Model Toward Unlimited Modalities","code":"https://github.com/modelscope/modelscope","n_code_links":2,"syntology":{"n_ran":2,"n_unverified":5,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":3,"model":"BLIP-2 (ViT-L, fine-tuned)","metrics":{"Recall@1":"83.5","Recall@10":"98.0","Recall@5":"96.0"},"uses_additional_data":false,"paper_date":"2023-01-30","paper":"/paper/blip-2-bootstrapping-language-image-pre","paper_url":"https://arxiv.org/abs/2301.12597v3","paper_title":"BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models","code":"https://github.com/huggingface/transformers","n_code_links":17,"syntology":{"n_ran":4,"n_unverified":4,"n_samples":8,"n_pointer_only_licence":1}},{"rank_in_archive_order":4,"model":"IAIS","metrics":{"Recall@1":"67.78","Recall@10":"94.48","Recall@5":"89.7"},"uses_additional_data":false,"paper_date":"2021-05-28","paper":"/paper/learning-relation-alignment-for-calibrated","paper_url":"https://arxiv.org/abs/2105.13868v2","paper_title":"Learning Relation Alignment for Calibrated Cross-modal Retrieval","code":"https://github.com/lancopku/IAIS","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":1,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":5,"model":"CLIP (zero-shot)","metrics":{"Recall@1":"58.4","Recall@10":"88.1","Recall@5":"81.5"},"uses_additional_data":false,"paper_date":"2021-02-26","paper":"/paper/learning-transferable-visual-models-from","paper_url":"https://arxiv.org/abs/2103.00020v1","paper_title":"Learning Transferable Visual Models From Natural Language Supervision","code":"https://github.com/openai/CLIP","n_code_links":82,"syntology":{"n_ran":16,"n_unverified":4,"n_samples":20,"n_pointer_only_licence":16}},{"rank_in_archive_order":6,"model":"FLAVA (ViT-B, zero-shot)","metrics":{"Recall@1":"42.74","Recall@5":"76.76"},"uses_additional_data":false,"paper_date":"2021-12-08","paper":"/paper/flava-a-foundational-language-and-vision","paper_url":"https://arxiv.org/abs/2112.04482v3","paper_title":"FLAVA: A Foundational Language And Vision Alignment Model","code":"https://github.com/facebookresearch/multimodal","n_code_links":4,"syntology":null},{"rank_in_archive_order":7,"model":"Oscar","metrics":{"Recall@10":"99.8"},"uses_additional_data":false,"paper_date":"2020-04-13","paper":"/paper/oscar-object-semantics-aligned-pre-training","paper_url":"https://arxiv.org/abs/2004.06165v5","paper_title":"Oscar: Object-Semantics Aligned Pre-training for Vision-Language Tasks","code":"https://github.com/rmokady/clip_prefix_caption","n_code_links":4,"syntology":{"n_ran":11,"n_unverified":12,"n_samples":23,"n_pointer_only_licence":0}},{"rank_in_archive_order":8,"model":"Unicoder-VL","metrics":{"Recall@10":"97.2"},"uses_additional_data":false,"paper_date":"2019-08-16","paper":"/paper/unicoder-vl-a-universal-encoder-for-vision","paper_url":"https://arxiv.org/abs/1908.06066v3","paper_title":"Unicoder-VL: A Universal Encoder for Vision and Language by Cross-modal Pre-training","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":9,"model":"DVSA","metrics":{"Recall@10":"74.8"},"uses_additional_data":false,"paper_date":"2014-12-07","paper":"/paper/deep-visual-semantic-alignments-for","paper_url":"http://arxiv.org/abs/1412.2306v2","paper_title":"Deep Visual-Semantic Alignments for Generating Image Descriptions","code":"https://github.com/VinitSR7/Image-Caption-Generation","n_code_links":4,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":0}}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,264 of the 9,581 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9581,"papers_checked":6264,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":3316},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-24T18:15:14+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":7,"rows_with_any_sample_ran":6,"distinct_papers_with_graph_line":6,"distinct_papers_with_any_sample_ran":5,"samples_over_distinct_papers":{"n_ran":34,"n_unverified":27,"n_samples":61,"n_pointer_only_licence":17,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":38,"n_unverified":31,"n_samples":69,"n_pointer_only_licence":18,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}