{"url":"/sota/image-retrieval-on-fashion-iq","task":{"name":"Image Retrieval","url":"/task/image-retrieval","note":null},"dataset":{"name":"Fashion IQ","url":"/dataset/fashion-iq"},"category":"Computer Vision","categories":["Computer Vision"],"category_note":null,"description":"**Image Retrieval** is a fundamental and long-standing computer vision task that involves finding images similar to a given query from a large database. It is often considered a form of fine-grained, instance-level classification. The task is integral to image recognition alongside [classification](/task/image-classification) and [cross-modal retrieval](/task/cross-modal-retrieva). By leveraging visual similarity and other criteria, image retrieval enables users to efficiently discover relevant images, making it a crucial tool in applications such as search and recommendation.\r\n\r\n<span class=\"description-source\">[Extending CLIP for Category-to-image Retrieval in E-commerce](https://arxiv.org/abs/2112.11294)</span>\r\n\r\n<span style=\"color:grey; opacity: 0.6\">( Image credit: [DELF](https://github.com/tensorflow/models/tree/master/research/delf) )</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@10+Recall@50)/2","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@10+Recall@50)/2":"higher","Recall@10":"higher"}},"counts":{"rows":22,"rows_with_code":17,"rows_with_paper_page":21,"rows_dated":21,"rows_using_additional_data":1},"rows":[{"rank_in_archive_order":1,"model":"DQU-CIR","metrics":{"(Recall@10+Recall@50)/2":"71.77","Recall@10":"61.97"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":2,"model":"TMCIR","metrics":{"(Recall@10+Recall@50)/2":"66.56","Recall@10":"56.57"},"uses_additional_data":false,"paper_date":"2025-04-15","paper":"/paper/tmcir-token-merge-benefits-composed-image","paper_url":"https://arxiv.org/abs/2504.10995v1","paper_title":"TMCIR: Token Merge Benefits Composed Image Retrieval","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":3,"model":"SPN4CIR (SPRC)","metrics":{"(Recall@10+Recall@50)/2":"66.41","Recall@10":"56.37"},"uses_additional_data":false,"paper_date":"2024-04-17","paper":"/paper/improving-composed-image-retrieval-via","paper_url":"https://arxiv.org/abs/2404.11317v2","paper_title":"Improving Composed Image Retrieval via Contrastive Learning with Scaling Positives and Negatives","code":"https://github.com/BUAADreamer/SPN4CIR","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":1,"n_samples":4,"n_pointer_only_licence":0}},{"rank_in_archive_order":4,"model":"SPRC","metrics":{"(Recall@10+Recall@50)/2":"64.85","Recall@10":"54.92"},"uses_additional_data":true,"paper_date":"2023-10-09","paper":"/paper/sentence-level-prompts-benefit-composed-image","paper_url":"https://arxiv.org/abs/2310.05473v1","paper_title":"Sentence-level Prompts Benefit Composed Image Retrieval","code":"https://github.com/chunmeifeng/sprc","n_code_links":1,"syntology":{"n_ran":7,"n_unverified":1,"n_samples":8,"n_pointer_only_licence":8}},{"rank_in_archive_order":5,"model":"Candidate Set Re-ranking","metrics":{"(Recall@10+Recall@50)/2":"62.15","Recall@10":"51.17"},"uses_additional_data":false,"paper_date":"2023-05-25","paper":"/paper/candidate-set-re-ranking-for-composed-image","paper_url":"https://arxiv.org/abs/2305.16304v3","paper_title":"Candidate Set Re-ranking for Composed Image Retrieval with Dual Multi-modal Encoder","code":"https://github.com/Cuberick-Orion/Bi-Blip4CIR","n_code_links":2,"syntology":{"n_ran":4,"n_unverified":2,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":6,"model":"RUTIR (BLIP B/16)","metrics":{"(Recall@10+Recall@50)/2":"61.32"},"uses_additional_data":false,"paper_date":"2023-08-16","paper":"/paper/ranking-aware-uncertainty-for-text-guided","paper_url":"https://arxiv.org/abs/2308.08131v1","paper_title":"Ranking-aware Uncertainty for Text-guided Image Retrieval","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":7,"model":"CASE","metrics":{"(Recall@10+Recall@50)/2":"59.73","Recall@10":"48.79"},"uses_additional_data":false,"paper_date":"2023-03-16","paper":"/paper/data-roaming-and-early-fusion-for-composed","paper_url":"https://arxiv.org/abs/2303.09429v2","paper_title":"Data Roaming and Quality Assessment for Composed Image Retrieval","code":"https://github.com/levymsn/LaSCo","n_code_links":1,"syntology":null},{"rank_in_archive_order":8,"model":"CaLa","metrics":{"(Recall@10+Recall@50)/2":"57.96","Recall@10":"46.69"},"uses_additional_data":false,"paper_date":"2024-05-29","paper":"/paper/cala-complementary-association-learning-for","paper_url":"https://arxiv.org/abs/2405.19149v2","paper_title":"CaLa: Complementary Association Learning for Augmenting Composed Image Retrieval","code":"https://github.com/chiangsonw/cala","n_code_links":1,"syntology":null},{"rank_in_archive_order":9,"model":"BLIP4CIR+Bi","metrics":{"(Recall@10+Recall@50)/2":"55.4"},"uses_additional_data":false,"paper_date":"2023-03-29","paper":"/paper/bi-directional-training-for-composed-image","paper_url":"https://arxiv.org/abs/2303.16604v2","paper_title":"Bi-directional Training for Composed Image Retrieval via Text Prompt Learning","code":"https://github.com/Cuberick-Orion/Bi-Blip4CIR","n_code_links":1,"syntology":null},{"rank_in_archive_order":10,"model":"CLIP4Cir (v3)","metrics":{"(Recall@10+Recall@50)/2":"55.36"},"uses_additional_data":false,"paper_date":"2023-08-22","paper":"/paper/composed-image-retrieval-using-contrastive","paper_url":"https://arxiv.org/abs/2308.11485v1","paper_title":"Composed Image Retrieval using Contrastive Learning and Task-oriented CLIP-based Features","code":"https://github.com/ABaldrati/CLIP4Cir","n_code_links":2,"syntology":{"n_ran":4,"n_unverified":1,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":11,"model":"RUTIR (CLIP ResNet50)","metrics":{"(Recall@10+Recall@50)/2":"55.27"},"uses_additional_data":false,"paper_date":"2023-08-16","paper":"/paper/ranking-aware-uncertainty-for-text-guided","paper_url":"https://arxiv.org/abs/2308.08131v1","paper_title":"Ranking-aware Uncertainty for Text-guided Image Retrieval","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":12,"model":"Css-Net","metrics":{"(Recall@10+Recall@50)/2":"51.34"},"uses_additional_data":false,"paper_date":"2023-06-03","paper":"/paper/relieving-triplet-ambiguity-consensus-network","paper_url":"https://arxiv.org/abs/2306.02092v2","paper_title":"Collaborative Group: Composed Image Retrieval via Consensus Learning from Noisy Annotations","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":13,"model":"MUR (4*ResNet50)","metrics":{"(Recall@10+Recall@50)/2":"50.61"},"uses_additional_data":false,"paper_date":"2022-11-14","paper":"/paper/composed-image-retrieval-with-text-feedback","paper_url":"https://arxiv.org/abs/2211.07394v6","paper_title":"Composed Image Retrieval with Text Feedback via Multi-grained Uncertainty Regularization","code":"https://github.com/Monoxide-Chen/uncertainty_retrieval","n_code_links":1,"syntology":{"n_ran":7,"n_unverified":3,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":14,"model":"CLIP4Cir (v2)","metrics":{"(Recall@10+Recall@50)/2":"50.03"},"uses_additional_data":false,"paper_date":"2022-06-19","paper":"/paper/conditioned-and-composed-image-retrieval","paper_url":"https://openaccess.thecvf.com/content/CVPR2022W/ODRUM/html/Baldrati_Conditioned_and_Composed_Image_Retrieval_Combining_and_Partially_Fine-Tuning_CLIP-Based_CVPRW_2022_paper.html","paper_title":"Conditioned and Composed Image Retrieval Combining and Partially Fine-Tuning CLIP-Based Features","code":"https://github.com/ABaldrati/CLIP4Cir","n_code_links":2,"syntology":null},{"rank_in_archive_order":15,"model":"MUR","metrics":{"(Recall@10+Recall@50)/2":"47.28"},"uses_additional_data":false,"paper_date":"2022-11-14","paper":"/paper/composed-image-retrieval-with-text-feedback","paper_url":"https://arxiv.org/abs/2211.07394v6","paper_title":"Composed Image Retrieval with Text Feedback via Multi-grained Uncertainty Regularization","code":"https://github.com/Monoxide-Chen/uncertainty_retrieval","n_code_links":1,"syntology":{"n_ran":7,"n_unverified":3,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":16,"model":"CLIP4Cir","metrics":{"(Recall@10+Recall@50)/2":"47.21"},"uses_additional_data":false,"paper_date":"2022-01-01","paper":"/paper/effective-conditioned-and-composed-image","paper_url":"http://openaccess.thecvf.com//content/CVPR2022/html/Baldrati_Effective_Conditioned_and_Composed_Image_Retrieval_Combining_CLIP-Based_Features_CVPR_2022_paper.html","paper_title":"Effective Conditioned and Composed Image Retrieval Combining CLIP-Based Features","code":"https://github.com/ABaldrati/CLIP4Cir","n_code_links":2,"syntology":null},{"rank_in_archive_order":17,"model":"MMRet-MLLM","metrics":{"(Recall@10+Recall@50)/2":"46.1","Recall@10":"35.6"},"uses_additional_data":false,"paper_date":"2024-12-19","paper":"/paper/megapairs-massive-data-synthesis-for","paper_url":"https://arxiv.org/abs/2412.14475v1","paper_title":"MegaPairs: Massive Data Synthesis For Universal Multimodal Retrieval","code":"https://github.com/VectorSpaceLab/MegaPairs","n_code_links":1,"syntology":null},{"rank_in_archive_order":18,"model":"RTIC-GCN","metrics":{"(Recall@10+Recall@50)/2":"40.64"},"uses_additional_data":false,"paper_date":"2021-04-07","paper":"/paper/rtic-residual-learning-for-text-and-image","paper_url":"https://arxiv.org/abs/2104.03015v3","paper_title":"RTIC: Residual Learning for Text and Image Composition using Graph Convolutional Network","code":"https://github.com/brandonhanx/compfashion","n_code_links":2,"syntology":null},{"rank_in_archive_order":19,"model":"CoSMo","metrics":{"(Recall@10+Recall@50)/2":"39.45"},"uses_additional_data":false,"paper_date":"2021-06-19","paper":"/paper/cosmo-content-style-modulation-for-image","paper_url":"http://openaccess.thecvf.com//content/CVPR2021/html/Lee_CoSMo_Content-Style_Modulation_for_Image_Retrieval_With_Text_Feedback_CVPR_2021_paper.html","paper_title":"CoSMo: Content-Style Modulation for Image Retrieval With Text Feedback","code":"https://github.com/postBG/CosMo.pytorch","n_code_links":1,"syntology":null},{"rank_in_archive_order":20,"model":"CurlingNet","metrics":{"(Recall@10+Recall@50)/2":"38.45"},"uses_additional_data":false,"paper_date":"2020-03-27","paper":"/paper/curlingnet-compositional-learning-between","paper_url":"https://arxiv.org/abs/2003.12299v2","paper_title":"CurlingNet: Compositional Learning between Images and Text for Fashion IQ Data","code":"https://github.com/nashory/rtic-gcn-pytorch","n_code_links":1,"syntology":null},{"rank_in_archive_order":21,"model":"VAL w/ GloVe","metrics":{"(Recall@10+Recall@50)/2":"35.38"},"uses_additional_data":false,"paper_date":"2020-06-01","paper":"/paper/image-search-with-text-feedback-by","paper_url":"http://openaccess.thecvf.com/content_CVPR_2020/html/Chen_Image_Search_With_Text_Feedback_by_Visiolinguistic_Attention_Learning_CVPR_2020_paper.html","paper_title":"Image Search With Text Feedback by Visiolinguistic Attention Learning","code":"https://github.com/yanbeic/VAL","n_code_links":1,"syntology":null},{"rank_in_archive_order":22,"model":"ComposeAE","metrics":{"(Recall@10+Recall@50)/2":"20.6"},"uses_additional_data":false,"paper_date":"2020-06-19","paper":"/paper/compositional-learning-of-image-text-query","paper_url":"https://arxiv.org/abs/2006.11149v3","paper_title":"Compositional Learning of Image-Text Query for Image Retrieval","code":"https://github.com/ecom-research/ComposeAE","n_code_links":1,"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 6,885 of the 9,623 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":9623,"papers_checked":6885,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2737},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-25T09:33:49+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":25,"n_unverified":9,"n_samples":34,"n_pointer_only_licence":8,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":32,"n_unverified":12,"n_samples":44,"n_pointer_only_licence":8,"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"}}}