{"url":"/sota/video-question-answering-on-tvbench","task":{"name":"Video Question Answering","url":"/task/video-question-answering","note":null},"dataset":{"name":"TVBench","url":"/dataset/tvbench"},"category":"Computer Vision","categories":["Computer Vision","Reasoning"],"category_note":null,"description":null,"description_from":null,"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":["Average Accuracy"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Average Accuracy":"higher"}},"counts":{"rows":28,"rows_with_code":22,"rows_with_paper_page":28,"rows_dated":28,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"Seed1.5-VL thinking","metrics":{"Average Accuracy":"63.6"},"uses_additional_data":false,"paper_date":"2025-05-11","paper":"/paper/seed1-5-vl-technical-report","paper_url":"https://arxiv.org/abs/2505.07062v1","paper_title":"Seed1.5-VL Technical Report","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":2,"model":"PLM-8B","metrics":{"Average Accuracy":"63.5"},"uses_additional_data":false,"paper_date":"2025-04-17","paper":"/paper/perceptionlm-open-access-data-and-models-for","paper_url":"https://arxiv.org/abs/2504.13180v1","paper_title":"PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding","code":"https://github.com/facebookresearch/perception_models","n_code_links":1,"syntology":null},{"rank_in_archive_order":3,"model":"Seed1.5-VL","metrics":{"Average Accuracy":"61.5"},"uses_additional_data":false,"paper_date":"2025-05-11","paper":"/paper/seed1-5-vl-technical-report","paper_url":"https://arxiv.org/abs/2505.07062v1","paper_title":"Seed1.5-VL Technical Report","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":4,"model":"V-JEPA 2 ViT-g 8B","metrics":{"Average Accuracy":"60.6"},"uses_additional_data":false,"paper_date":"2025-06-11","paper":"/paper/v-jepa-2-self-supervised-video-models-enable","paper_url":"https://arxiv.org/abs/2506.09985v1","paper_title":"V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning","code":"https://github.com/facebookresearch/vjepa2","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":8,"n_samples":8,"n_pointer_only_licence":8}},{"rank_in_archive_order":5,"model":"PLM-3B","metrics":{"Average Accuracy":"58.9"},"uses_additional_data":false,"paper_date":"2025-04-17","paper":"/paper/perceptionlm-open-access-data-and-models-for","paper_url":"https://arxiv.org/abs/2504.13180v1","paper_title":"PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding","code":"https://github.com/facebookresearch/perception_models","n_code_links":1,"syntology":null},{"rank_in_archive_order":6,"model":"RRPO","metrics":{"Average Accuracy":"56.5"},"uses_additional_data":false,"paper_date":"2025-04-16","paper":"/paper/self-alignment-of-large-video-language-models","paper_url":"https://arxiv.org/abs/2504.12083v1","paper_title":"Self-alignment of Large Video Language Models with Refined Regularized Preference Optimization","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":7,"model":"Tarsier-34B","metrics":{"Average Accuracy":"55.5"},"uses_additional_data":false,"paper_date":"2024-06-30","paper":"/paper/tarsier-recipes-for-training-and-evaluating-1","paper_url":"https://arxiv.org/abs/2407.00634v2","paper_title":"Tarsier: Recipes for Training and Evaluating Large Video Description Models","code":"https://github.com/bytedance/tarsier","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":8,"model":"Tarsier2-7B","metrics":{"Average Accuracy":"54.7"},"uses_additional_data":false,"paper_date":"2025-01-14","paper":"/paper/tarsier2-advancing-large-vision-language","paper_url":"https://arxiv.org/abs/2501.07888v3","paper_title":"Tarsier2: Advancing Large Vision-Language Models from Detailed Video Description to Comprehensive Video Understanding","code":"https://github.com/bytedance/tarsier","n_code_links":1,"syntology":{"n_ran":4,"n_unverified":0,"n_samples":4,"n_pointer_only_licence":4}},{"rank_in_archive_order":9,"model":"Qwen2-VL-72B","metrics":{"Average Accuracy":"52.7"},"uses_additional_data":false,"paper_date":"2024-09-18","paper":"/paper/qwen2-vl-enhancing-vision-language-model-s","paper_url":"https://arxiv.org/abs/2409.12191v2","paper_title":"Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution","code":"https://github.com/qwenlm/qwen2-vl","n_code_links":8,"syntology":{"n_ran":8,"n_unverified":4,"n_samples":12,"n_pointer_only_licence":0}},{"rank_in_archive_order":10,"model":"IXC-2.5 7B","metrics":{"Average Accuracy":"51.6"},"uses_additional_data":false,"paper_date":"2024-07-03","paper":"/paper/internlm-xcomposer-2-5-a-versatile-large","paper_url":"https://arxiv.org/abs/2407.03320v1","paper_title":"InternLM-XComposer-2.5: A Versatile Large Vision Language Model Supporting Long-Contextual Input and Output","code":"https://github.com/internlm/internlm-xcomposer","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":11,"model":"Aria","metrics":{"Average Accuracy":"51.0"},"uses_additional_data":false,"paper_date":"2024-10-08","paper":"/paper/aria-an-open-multimodal-native-mixture-of","paper_url":"https://arxiv.org/abs/2410.05993v4","paper_title":"Aria: An Open Multimodal Native Mixture-of-Experts Model","code":"https://github.com/rhymes-ai/aria","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":12,"model":"PLM-1B","metrics":{"Average Accuracy":"50.4"},"uses_additional_data":false,"paper_date":"2025-04-17","paper":"/paper/perceptionlm-open-access-data-and-models-for","paper_url":"https://arxiv.org/abs/2504.13180v1","paper_title":"PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding","code":"https://github.com/facebookresearch/perception_models","n_code_links":1,"syntology":null},{"rank_in_archive_order":13,"model":"LLaVA-Video 72B","metrics":{"Average Accuracy":"50.0"},"uses_additional_data":false,"paper_date":"2024-10-03","paper":"/paper/video-instruction-tuning-with-synthetic-data","paper_url":"https://arxiv.org/abs/2410.02713v2","paper_title":"Video Instruction Tuning With Synthetic Data","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":14,"model":"VideoLLaMA2 72B","metrics":{"Average Accuracy":"48.4"},"uses_additional_data":false,"paper_date":"2024-06-11","paper":"/paper/videollama-2-advancing-spatial-temporal","paper_url":"https://arxiv.org/abs/2406.07476v3","paper_title":"VideoLLaMA 2: Advancing Spatial-Temporal Modeling and Audio Understanding in Video-LLMs","code":"https://github.com/damo-nlp-sg/videollama2","n_code_links":3,"syntology":{"n_ran":7,"n_unverified":10,"n_samples":17,"n_pointer_only_licence":0}},{"rank_in_archive_order":15,"model":"Gemini 1.5 Pro","metrics":{"Average Accuracy":"47.6"},"uses_additional_data":false,"paper_date":"2024-03-08","paper":"/paper/gemini-1-5-unlocking-multimodal-understanding","paper_url":"https://arxiv.org/abs/2403.05530v5","paper_title":"Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context","code":"https://github.com/dlvuldet/primevul","n_code_links":1,"syntology":null},{"rank_in_archive_order":16,"model":"Tarsier-7B","metrics":{"Average Accuracy":"46.9"},"uses_additional_data":false,"paper_date":"2024-06-30","paper":"/paper/tarsier-recipes-for-training-and-evaluating-1","paper_url":"https://arxiv.org/abs/2407.00634v2","paper_title":"Tarsier: Recipes for Training and Evaluating Large Video Description Models","code":"https://github.com/bytedance/tarsier","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":17,"model":"LLaVA-Video 7B","metrics":{"Average Accuracy":"45.6"},"uses_additional_data":false,"paper_date":"2024-10-03","paper":"/paper/video-instruction-tuning-with-synthetic-data","paper_url":"https://arxiv.org/abs/2410.02713v2","paper_title":"Video Instruction Tuning With Synthetic Data","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":18,"model":"Qwen2-VL-7B","metrics":{"Average Accuracy":"43.8"},"uses_additional_data":false,"paper_date":"2024-09-18","paper":"/paper/qwen2-vl-enhancing-vision-language-model-s","paper_url":"https://arxiv.org/abs/2409.12191v2","paper_title":"Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution","code":"https://github.com/qwenlm/qwen2-vl","n_code_links":8,"syntology":{"n_ran":8,"n_unverified":4,"n_samples":12,"n_pointer_only_licence":0}},{"rank_in_archive_order":19,"model":"VideoLLaMA2 7B","metrics":{"Average Accuracy":"42.9"},"uses_additional_data":false,"paper_date":"2024-06-11","paper":"/paper/videollama-2-advancing-spatial-temporal","paper_url":"https://arxiv.org/abs/2406.07476v3","paper_title":"VideoLLaMA 2: Advancing Spatial-Temporal Modeling and Audio Understanding in Video-LLMs","code":"https://github.com/damo-nlp-sg/videollama2","n_code_links":3,"syntology":{"n_ran":7,"n_unverified":10,"n_samples":17,"n_pointer_only_licence":0}},{"rank_in_archive_order":20,"model":"PLLaVA-34B","metrics":{"Average Accuracy":"42.3"},"uses_additional_data":false,"paper_date":"2024-04-25","paper":"/paper/pllava-parameter-free-llava-extension-from-1","paper_url":"https://arxiv.org/abs/2404.16994v2","paper_title":"PLLaVA : Parameter-free LLaVA Extension from Images to Videos for Video Dense Captioning","code":"https://github.com/magic-research/PLLaVA","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":21,"model":"mPLUG-Owl3","metrics":{"Average Accuracy":"42.2"},"uses_additional_data":false,"paper_date":"2024-08-09","paper":"/paper/mplug-owl3-towards-long-image-sequence","paper_url":"https://arxiv.org/abs/2408.04840v2","paper_title":"mPLUG-Owl3: Towards Long Image-Sequence Understanding in Multi-Modal Large Language Models","code":"https://github.com/x-plug/mplug-owl","n_code_links":1,"syntology":null},{"rank_in_archive_order":22,"model":"VideoLLaMA2.1","metrics":{"Average Accuracy":"42.1"},"uses_additional_data":false,"paper_date":"2024-06-11","paper":"/paper/videollama-2-advancing-spatial-temporal","paper_url":"https://arxiv.org/abs/2406.07476v3","paper_title":"VideoLLaMA 2: Advancing Spatial-Temporal Modeling and Audio Understanding in Video-LLMs","code":"https://github.com/damo-nlp-sg/videollama2","n_code_links":3,"syntology":{"n_ran":7,"n_unverified":10,"n_samples":17,"n_pointer_only_licence":0}},{"rank_in_archive_order":23,"model":"VideoGPT+","metrics":{"Average Accuracy":"41.7"},"uses_additional_data":false,"paper_date":"2024-06-13","paper":"/paper/videogpt-integrating-image-and-video-encoders","paper_url":"https://arxiv.org/abs/2406.09418v1","paper_title":"VideoGPT+: Integrating Image and Video Encoders for Enhanced Video Understanding","code":"https://github.com/mbzuai-oryx/videogpt-plus","n_code_links":1,"syntology":{"n_ran":6,"n_unverified":2,"n_samples":8,"n_pointer_only_licence":8}},{"rank_in_archive_order":24,"model":"GPT4o 8 frames","metrics":{"Average Accuracy":"39.9"},"uses_additional_data":false,"paper_date":"2024-10-25","paper":"/paper/gpt-4o-system-card","paper_url":"https://arxiv.org/abs/2410.21276v1","paper_title":"GPT-4o System Card","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":25,"model":"PLLaVA-13B","metrics":{"Average Accuracy":"36.4"},"uses_additional_data":false,"paper_date":"2024-04-25","paper":"/paper/pllava-parameter-free-llava-extension-from-1","paper_url":"https://arxiv.org/abs/2404.16994v2","paper_title":"PLLaVA : Parameter-free LLaVA Extension from Images to Videos for Video Dense Captioning","code":"https://github.com/magic-research/PLLaVA","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":2}},{"rank_in_archive_order":26,"model":"ST-LLM","metrics":{"Average Accuracy":"35.7"},"uses_additional_data":false,"paper_date":"2024-03-30","paper":"/paper/st-llm-large-language-models-are-effective-1","paper_url":"https://arxiv.org/abs/2404.00308v1","paper_title":"ST-LLM: Large Language Models Are Effective Temporal Learners","code":"https://github.com/TencentARC/ST-LLM","n_code_links":1,"syntology":{"n_ran":7,"n_unverified":4,"n_samples":11,"n_pointer_only_licence":3}},{"rank_in_archive_order":27,"model":"VideoChat2","metrics":{"Average Accuracy":"35.0"},"uses_additional_data":false,"paper_date":"2023-11-28","paper":"/paper/mvbench-a-comprehensive-multi-modal-video","paper_url":"https://arxiv.org/abs/2311.17005v4","paper_title":"MVBench: A Comprehensive Multi-modal Video Understanding Benchmark","code":"https://github.com/opengvlab/ask-anything","n_code_links":3,"syntology":{"n_ran":7,"n_unverified":3,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":28,"model":"PLLaVA-7B","metrics":{"Average Accuracy":"34.9"},"uses_additional_data":false,"paper_date":"2024-04-25","paper":"/paper/pllava-parameter-free-llava-extension-from-1","paper_url":"https://arxiv.org/abs/2404.16994v2","paper_title":"PLLaVA : Parameter-free LLaVA Extension from Images to Videos for Video Dense Captioning","code":"https://github.com/magic-research/PLLaVA","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":2}}],"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":17,"rows_with_any_sample_ran":13,"distinct_papers_with_graph_line":11,"distinct_papers_with_any_sample_ran":9,"samples_over_distinct_papers":{"n_ran":44,"n_unverified":33,"n_samples":77,"n_pointer_only_licence":27,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":68,"n_unverified":61,"n_samples":129,"n_pointer_only_licence":31,"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"}}}