{"url":"/sota/visual-question-answering-on-ok-vqa","task":{"name":"Visual Question Answering (VQA)","url":"/task/visual-question-answering","note":null},"dataset":{"name":"OK-VQA","url":"/dataset/ok-vqa"},"category":"Computer Vision","categories":["Computer Vision","Natural Language Processing"],"category_note":null,"description":"**Visual Question Answering (VQA)** is a task in computer vision that involves answering questions about an image. The goal of VQA is to teach machines to understand the content of an image and answer questions about it in natural language.\r\n\r\nImage Source: [visualqa.org](https://visualqa.org/)","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":["Accuracy","Exact Match (EM)","Recall@5"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Accuracy":"higher","Exact Match (EM)":null,"Recall@5":"higher"}},"counts":{"rows":37,"rows_with_code":31,"rows_with_paper_page":37,"rows_dated":37,"rows_using_additional_data":1},"rows":[{"rank_in_archive_order":1,"model":"PaLI-X-VPD","metrics":{"Accuracy":"66.8"},"uses_additional_data":false,"paper_date":"2023-12-05","paper":"/paper/visual-program-distillation-distilling-tools","paper_url":"https://arxiv.org/abs/2312.03052v2","paper_title":"Visual Program Distillation: Distilling Tools and Programmatic Reasoning into Vision-Language Models","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":2,"model":"PaLM-E-562B","metrics":{"Accuracy":"66.1"},"uses_additional_data":false,"paper_date":"2023-03-06","paper":"/paper/palm-e-an-embodied-multimodal-language-model","paper_url":"https://arxiv.org/abs/2303.03378v1","paper_title":"PaLM-E: An Embodied Multimodal Language Model","code":"https://github.com/kyegomez/PALM-E","n_code_links":2,"syntology":null},{"rank_in_archive_order":3,"model":"PaLI-X (Single-task FT)","metrics":{"Accuracy":"66.1"},"uses_additional_data":false,"paper_date":"2023-05-29","paper":"/paper/pali-x-on-scaling-up-a-multilingual-vision","paper_url":"https://arxiv.org/abs/2305.18565v1","paper_title":"PaLI-X: On Scaling up a Multilingual Vision and Language Model","code":"https://github.com/kyegomez/PALI","n_code_links":2,"syntology":{"n_ran":6,"n_unverified":1,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":4,"model":"PaLI 17B","metrics":{"Accuracy":"64.5"},"uses_additional_data":false,"paper_date":"2022-09-14","paper":"/paper/pali-a-jointly-scaled-multilingual-language","paper_url":"https://arxiv.org/abs/2209.06794v4","paper_title":"PaLI: A Jointly-Scaled Multilingual Language-Image Model","code":"https://github.com/google-research/big_vision","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":2,"n_samples":4,"n_pointer_only_licence":0}},{"rank_in_archive_order":5,"model":"Prophet","metrics":{"Accuracy":"62.5"},"uses_additional_data":false,"paper_date":"2023-03-03","paper":"/paper/prompting-large-language-models-with-answer","paper_url":"https://arxiv.org/abs/2303.01903v4","paper_title":"Prophet: Prompting Large Language Models with Complementary Answer Heuristics for Knowledge-based Visual Question Answering","code":"https://github.com/milvlg/prophet","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":8,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":6,"model":"RA-VQA-v2 (BLIP 2)","metrics":{"Accuracy":"62.08","Exact Match (EM)":"62.01","Recall@5":"89.32"},"uses_additional_data":false,"paper_date":"2023-09-29","paper":"/paper/fine-grained-late-interaction-multi-modal-1","paper_url":"https://arxiv.org/abs/2309.17133v2","paper_title":"Fine-grained Late-interaction Multi-modal Retrieval for Retrieval Augmented Visual Question Answering","code":"https://github.com/linweizhedragon/retrieval-augmented-visual-question-answering","n_code_links":1,"syntology":{"n_ran":8,"n_unverified":0,"n_samples":8,"n_pointer_only_licence":8}},{"rank_in_archive_order":7,"model":"A Simple Baseline for KB-VQA","metrics":{"Accuracy":"61.2"},"uses_additional_data":false,"paper_date":"2023-10-20","paper":"/paper/a-simple-baseline-for-knowledge-based-visual","paper_url":"https://arxiv.org/abs/2310.13570v2","paper_title":"A Simple Baseline for Knowledge-Based Visual Question Answering","code":null,"n_code_links":0,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":8,"model":"PromptCap","metrics":{"Accuracy":"60.4"},"uses_additional_data":false,"paper_date":"2022-11-15","paper":"/paper/promptcap-prompt-guided-task-aware-image","paper_url":"https://arxiv.org/abs/2211.09699v4","paper_title":"PromptCap: Prompt-Guided Task-Aware Image Captioning","code":"https://github.com/Yushi-Hu/PromptCap","n_code_links":1,"syntology":null},{"rank_in_archive_order":9,"model":"ReVeaL WIT + CC12M + Wikidata + VQA-2","metrics":{"Accuracy":"59.1"},"uses_additional_data":false,"paper_date":"2022-12-10","paper":"/paper/reveal-retrieval-augmented-visual-language","paper_url":"https://arxiv.org/abs/2212.05221v2","paper_title":"REVEAL: Retrieval-Augmented Visual-Language Pre-Training with Multi-Source Multimodal Knowledge Memory","code":"https://github.com/google-research/scenic/tree/main/scenic/projects/knowledge_visual_language","n_code_links":1,"syntology":null},{"rank_in_archive_order":10,"model":"Lyrics","metrics":{"Accuracy":"58.2"},"uses_additional_data":false,"paper_date":"2023-12-08","paper":"/paper/lyrics-boosting-fine-grained-language-vision","paper_url":"https://arxiv.org/abs/2312.05278v2","paper_title":"Lyrics: Boosting Fine-grained Language-Vision Alignment and Comprehension via Semantic-aware Visual Objects","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":11,"model":"REVIVE (Ensemble)","metrics":{"Accuracy":"58.0"},"uses_additional_data":false,"paper_date":"2022-06-02","paper":"/paper/revive-regional-visual-representation-matters","paper_url":"https://arxiv.org/abs/2206.01201v2","paper_title":"REVIVE: Regional Visual Representation Matters in Knowledge-Based Visual Question Answering","code":"https://github.com/yzleroy/revive","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":12,"model":"REVIVE (Single)","metrics":{"Accuracy":"56.6"},"uses_additional_data":false,"paper_date":"2022-06-02","paper":"/paper/revive-regional-visual-representation-matters","paper_url":"https://arxiv.org/abs/2206.01201v2","paper_title":"REVIVE: Regional Visual Representation Matters in Knowledge-Based Visual Question Answering","code":"https://github.com/yzleroy/revive","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":13,"model":"RA-VQA-v2 (T5-large)","metrics":{"Accuracy":"54.85"},"uses_additional_data":false,"paper_date":"2023-09-29","paper":"/paper/fine-grained-late-interaction-multi-modal-1","paper_url":"https://arxiv.org/abs/2309.17133v2","paper_title":"Fine-grained Late-interaction Multi-modal Retrieval for Retrieval Augmented Visual Question Answering","code":"https://github.com/linweizhedragon/retrieval-augmented-visual-question-answering","n_code_links":1,"syntology":{"n_ran":8,"n_unverified":0,"n_samples":8,"n_pointer_only_licence":8}},{"rank_in_archive_order":14,"model":"RA-VQA (T5-large)","metrics":{"Accuracy":"54.48","Exact Match (EM)":"59.41","Recall@5":"82.84"},"uses_additional_data":false,"paper_date":"2022-10-07","paper":"/paper/retrieval-augmented-visual-question-answering","paper_url":"https://arxiv.org/abs/2210.03809v2","paper_title":"Retrieval Augmented Visual Question Answering with Outside Knowledge","code":"https://github.com/linweizhedragon/retrieval-augmented-visual-question-answering","n_code_links":1,"syntology":null},{"rank_in_archive_order":15,"model":"VK-OOD","metrics":{"Accuracy":"52.4"},"uses_additional_data":false,"paper_date":"2023-02-11","paper":"/paper/differentiable-outlier-detection-enable","paper_url":"https://arxiv.org/abs/2302.05608v1","paper_title":"Differentiable Outlier Detection Enable Robust Deep Multimodal Analysis","code":"https://github.com/ellenzhuwang/VK_OOD","n_code_links":1,"syntology":null},{"rank_in_archive_order":16,"model":"VK-OOD","metrics":{"Accuracy":"52.4"},"uses_additional_data":false,"paper_date":"2023-09-21","paper":"/paper/implicit-differentiable-outlier-detection","paper_url":"https://openreview.net/forum?id=jooPcatnVF","paper_title":"Implicit Differentiable Outlier Detection Enable Robust Deep Multimodal Analysis","code":"https://github.com/ellenzhuwang/implicit_vkood","n_code_links":1,"syntology":null},{"rank_in_archive_order":17,"model":"RA-VQA-FrDPR (T5-large)","metrics":{"Accuracy":"51.22","Exact Match (EM)":"55.77","Recall@5":"81.25"},"uses_additional_data":false,"paper_date":"2022-10-07","paper":"/paper/retrieval-augmented-visual-question-answering","paper_url":"https://arxiv.org/abs/2210.03809v2","paper_title":"Retrieval Augmented Visual Question Answering with Outside Knowledge","code":"https://github.com/linweizhedragon/retrieval-augmented-visual-question-answering","n_code_links":1,"syntology":null},{"rank_in_archive_order":18,"model":"Flamingo80B","metrics":{"Accuracy":"50.6"},"uses_additional_data":false,"paper_date":"2022-04-29","paper":"/paper/flamingo-a-visual-language-model-for-few-shot-1","paper_url":"https://arxiv.org/abs/2204.14198v2","paper_title":"Flamingo: a Visual Language Model for Few-Shot Learning","code":"https://github.com/mlfoundations/open_flamingo","n_code_links":5,"syntology":{"n_ran":18,"n_unverified":6,"n_samples":24,"n_pointer_only_licence":7}},{"rank_in_archive_order":19,"model":"TRiG (T5-Large)","metrics":{"Accuracy":"50.50"},"uses_additional_data":false,"paper_date":"2022-01-01","paper":"/paper/transform-retrieve-generate-natural-language","paper_url":"http://openaccess.thecvf.com//content/CVPR2022/html/Gao_Transform-Retrieve-Generate_Natural_Language-Centric_Outside-Knowledge_Visual_Question_Answering_CVPR_2022_paper.html","paper_title":"Transform-Retrieve-Generate: Natural Language-Centric Outside-Knowledge Visual Question Answering","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":20,"model":"HYDRA","metrics":{"Accuracy":"48.6"},"uses_additional_data":false,"paper_date":"2024-03-19","paper":"/paper/hydra-a-hyper-agent-for-dynamic-compositional","paper_url":"https://arxiv.org/abs/2403.12884v2","paper_title":"HYDRA: A Hyper Agent for Dynamic Compositional Visual Reasoning","code":"https://github.com/ControlNet/HYDRA","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":3,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":21,"model":"PICa","metrics":{"Accuracy":"48.0"},"uses_additional_data":true,"paper_date":"2021-09-10","paper":"/paper/an-empirical-study-of-gpt-3-for-few-shot","paper_url":"https://arxiv.org/abs/2109.05014v2","paper_title":"An Empirical Study of GPT-3 for Few-Shot Knowledge-Based VQA","code":"https://github.com/microsoft/PICa","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":22,"model":"LaKo","metrics":{"Accuracy":"47.01"},"uses_additional_data":false,"paper_date":"2022-07-26","paper":"/paper/lako-knowledge-driven-visual-question","paper_url":"https://arxiv.org/abs/2207.12888v2","paper_title":"LaKo: Knowledge-driven Visual Question Answering via Late Knowledge-to-Text Injection","code":"https://github.com/hackerchenzhuo/LaKo","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":23,"model":"BLIP-2 ViT-G FlanT5 XXL (zero-shot)","metrics":{"Accuracy":"45.9"},"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":24,"model":"Flamingo9B","metrics":{"Accuracy":"44.7"},"uses_additional_data":false,"paper_date":"2022-04-29","paper":"/paper/flamingo-a-visual-language-model-for-few-shot-1","paper_url":"https://arxiv.org/abs/2204.14198v2","paper_title":"Flamingo: a Visual Language Model for Few-Shot Learning","code":"https://github.com/mlfoundations/open_flamingo","n_code_links":5,"syntology":{"n_ran":18,"n_unverified":6,"n_samples":24,"n_pointer_only_licence":7}},{"rank_in_archive_order":25,"model":"VLC-BERT","metrics":{"Accuracy":"43.1"},"uses_additional_data":false,"paper_date":"2022-10-24","paper":"/paper/vlc-bert-visual-question-answering-with","paper_url":"https://arxiv.org/abs/2210.13626v1","paper_title":"VLC-BERT: Visual Question Answering with Contextualized Commonsense Knowledge","code":"https://github.com/aditya10/vlc-bert","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":26,"model":"T5(Tan and Bansal, 2019) + Prefixes","metrics":{"Accuracy":"42.03"},"uses_additional_data":false,"paper_date":"2022-07-26","paper":"/paper/lako-knowledge-driven-visual-question","paper_url":"https://arxiv.org/abs/2207.12888v2","paper_title":"LaKo: Knowledge-driven Visual Question Answering via Late Knowledge-to-Text Injection","code":"https://github.com/hackerchenzhuo/LaKo","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":27,"model":"Flamingo3B","metrics":{"Accuracy":"41.2"},"uses_additional_data":false,"paper_date":"2022-04-29","paper":"/paper/flamingo-a-visual-language-model-for-few-shot-1","paper_url":"https://arxiv.org/abs/2204.14198v2","paper_title":"Flamingo: a Visual Language Model for Few-Shot Learning","code":"https://github.com/mlfoundations/open_flamingo","n_code_links":5,"syntology":{"n_ran":18,"n_unverified":6,"n_samples":24,"n_pointer_only_licence":7}},{"rank_in_archive_order":28,"model":"BLIP-2 ViT-G FlanT5 XL (zero-shot)","metrics":{"Accuracy":"40.7"},"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":29,"model":"BLIP-2 ViT-L FlanT5 XL (zero-shot)","metrics":{"Accuracy":"39.4"},"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":30,"model":"BLIP-2 ViT-G OPT 6.7B (zero-shot)","metrics":{"Accuracy":"36.4"},"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":31,"model":"PNP-VQA","metrics":{"Accuracy":"35.9"},"uses_additional_data":false,"paper_date":"2022-10-17","paper":"/paper/plug-and-play-vqa-zero-shot-vqa-by-conjoining","paper_url":"https://arxiv.org/abs/2210.08773v3","paper_title":"Plug-and-Play VQA: Zero-shot VQA by Conjoining Large Pretrained Models with Zero Training","code":"https://github.com/salesforce/lavis","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":32,"model":"BLIP-2 ViT-G OPT 2.7B (zero-shot)","metrics":{"Accuracy":"31.7"},"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":33,"model":"BLIP-2 ViT-L OPT 2.7B (zero-shot)","metrics":{"Accuracy":"30.2"},"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":34,"model":"FewVLM","metrics":{"Accuracy":"16.5"},"uses_additional_data":false,"paper_date":"2021-10-16","paper":"/paper/a-good-prompt-is-worth-millions-of-parameters","paper_url":"https://arxiv.org/abs/2110.08484v2","paper_title":"A Good Prompt Is Worth Millions of Parameters: Low-resource Prompt-based Learning for Vision-Language Models","code":"https://github.com/woojeongjin/fewvlm","n_code_links":1,"syntology":null},{"rank_in_archive_order":35,"model":"MetaLM","metrics":{"Accuracy":"11.4"},"uses_additional_data":false,"paper_date":"2022-06-13","paper":"/paper/language-models-are-general-purpose","paper_url":"https://arxiv.org/abs/2206.06336v1","paper_title":"Language Models are General-Purpose Interfaces","code":"https://github.com/microsoft/unilm","n_code_links":1,"syntology":null},{"rank_in_archive_order":36,"model":"VLKD(ViT-B/16)","metrics":{"Accuracy":"10.5"},"uses_additional_data":false,"paper_date":"2021-11-16","paper":"/paper/enabling-multimodal-generation-on-clip-via","paper_url":"https://openreview.net/forum?id=YTGg7kv8qIq","paper_title":"Enabling Multimodal Generation on CLIP via Vision-Language Knowledge Distillation","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":37,"model":"Frozen","metrics":{"Accuracy":" 5.9"},"uses_additional_data":false,"paper_date":"2021-06-25","paper":"/paper/multimodal-few-shot-learning-with-frozen","paper_url":"https://arxiv.org/abs/2106.13884v2","paper_title":"Multimodal Few-Shot Learning with Frozen Language Models","code":null,"n_code_links":0,"syntology":null}],"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 6,795 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":6795,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2785},"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":23,"rows_with_any_sample_ran":21,"distinct_papers_with_graph_line":13,"distinct_papers_with_any_sample_ran":11,"samples_over_distinct_papers":{"n_ran":52,"n_unverified":28,"n_samples":80,"n_pointer_only_licence":25,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":119,"n_unverified":60,"n_samples":179,"n_pointer_only_licence":55,"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"}}}