{"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/svitt-temporal-learning-of-sparse-video-text","title":"SViTT: Temporal Learning of Sparse Video-Text Transformers","arxiv_id":"2304.08809","date":"2023-04-18","proceeding":"CVPR 2023 1","authors":["Yi Li","Kyle Min","Subarna Tripathi","Nuno Vasconcelos"],"abstract":"Do video-text transformers learn to model temporal relationships across frames? Despite their immense capacity and the abundance of multimodal training data, recent work has revealed the strong tendency of video-text models towards frame-based spatial representations, while temporal reasoning remains largely unsolved. In this work, we identify several key challenges in temporal learning of video-text transformers: the spatiotemporal trade-off from limited network size; the curse of dimensionality for multi-frame modeling; and the diminishing returns of semantic information by extending clip length. Guided by these findings, we propose SViTT, a sparse video-text architecture that performs multi-frame reasoning with significantly lower cost than naive transformers with dense attention. Analogous to graph-based networks, SViTT employs two forms of sparsity: edge sparsity that limits the query-key communications between tokens in self-attention, and node sparsity that discards uninformative visual tokens. Trained with a curriculum which increases model sparsity with the clip length, SViTT outperforms dense transformer baselines on multiple video-text retrieval and question answering benchmarks, with a fraction of computational cost. Project page: http://svcl.ucsd.edu/projects/svitt.","url_abs":"https://arxiv.org/abs/2304.08809v1","url_pdf":"https://arxiv.org/pdf/2304.08809v1.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":"svitt-temporal-learning-of-sparse-video-text","repo_url":"https://github.com/jerryyli/svitt","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"question-answering","task_name":"Question Answering"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"text-retrieval","task_name":"Text Retrieval"},{"task_slug":"video-question-answering","task_name":"Video Question Answering"},{"task_slug":"video-text-retrieval","task_name":"Video-Text Retrieval"}],"methods":[{"method_slug":"clip","method_name":"CLIP"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/video-question-answering-on-agqa-2-0-balanced","task":"Video Question Answering","dataset":"AGQA 2.0 balanced","model":"SViTT","rank_in_archive_order":4,"of":8,"metrics":{"Average Accuracy":"52.7"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2304.08809","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}