{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T10:04:14Z","timestamp":1777889054354,"version":"3.51.4"},"reference-count":64,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,10,19]]},"DOI":"10.1109\/iccv51701.2025.02208","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"23784-23794","source":"Crossref","is-referenced-by-count":0,"title":["SparseVILA: Decoupling Visual Sparsity for Efficient VLM Inference"],"prefix":"10.1109","author":[{"given":"Samir","family":"Khaki","sequence":"first","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Junxian","family":"Guo","sequence":"additional","affiliation":[{"name":"MIT"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jiaming","family":"Tang","sequence":"additional","affiliation":[{"name":"MIT"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Shang","family":"Yang","sequence":"additional","affiliation":[{"name":"MIT"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yukang","family":"Chen","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Konstantinos N.","family":"Plataniotis","sequence":"additional","affiliation":[{"name":"University of Toronto"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yao","family":"Lu","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Song","family":"Han","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zhijian","family":"Liu","sequence":"additional","affiliation":[{"name":"NVIDIA"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","author":"Hasan","year":"2024","journal-title":"Hired: Attention-guided token dropping for efficient inference of high-resolution vision-language models in resourceconstrained environments"},{"key":"ref2","author":"Bai","year":"2023","journal-title":"Qwen-vl: A versatile vision-language model for understanding, localization, text reading, and beyond"},{"key":"ref3","author":"Bolya","year":"2022","journal-title":"Token merging: Your vit but faster"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73004-7_2"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00205"},{"key":"ref6","author":"Dao","year":"2023","journal-title":"Flashattention-2: Faster attention with better parallelism and work partitioning"},{"key":"ref7","author":"Dubey","year":"2024","journal-title":"The llama 3 herd of models"},{"key":"ref8","author":"Fu","year":"2024","journal-title":"Mme: A comprehensive evaluation benchmark for multimodal large language models"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.02245"},{"key":"ref10","author":"Fu","year":"2024","journal-title":"Mme-survey: A comprehensive survey on evaluation of multimodal 11 ms"},{"key":"ref11","author":"Fu","year":"2024","journal-title":"Lazyllm: Dynamic token pruning for efficient long context llm inference"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/ICCVW60793.2023.00085"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0040"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72643-9_13"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00686"},{"key":"ref16","author":"Kang","year":"2025","journal-title":"See what you are told: Visual attention sink in large multimodal models"},{"key":"ref17","article-title":"The need for speed: Pruning transformers with one recipe","volume-title":"In The Twelfth International Conference on Learning Representations","author":"Khaki","year":"2024"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1145\/3534678.3539260"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-20083-0_37"},{"key":"ref20","author":"Li","year":"2024","journal-title":"lavanext: Stronger 11 ms supercharge multimodal capabilities in the wild"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.20"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0722"},{"key":"ref23","author":"Liang","year":"2022","journal-title":"Not all patches are what you need: Expediting vision transformers via token reorganizations"},{"key":"ref24","author":"Lin","year":"2023","journal-title":"Vila: On pre-training for visual language models"},{"key":"ref25","author":"Lin","year":"2024","journal-title":"Awq: Activation-aware weight quantization for 11 m compression and acceleration"},{"key":"ref26","author":"Lin","year":"2024","journal-title":"Qserve: W4a8kv4 quantization and system co-design for efficient llm serving"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1516"},{"key":"ref28","author":"Liu","year":"2024","journal-title":"Llava-next: Improved reasoning, ocr, and world knowledge"},{"key":"ref29","author":"Liu","year":"2024","journal-title":"Kivi: A tuning-free asymmetric 2bit quantization for kv cache"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.00390"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.52202\/068431-0182"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.52202\/075280-0950"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.679"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.findings-acl.177"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1109\/WACV48630.2021.00225"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/WACV51458.2022.00264"},{"key":"ref37","author":"Radford","year":"2021","journal-title":"Learning transferable visual models from natural language supervision"},{"key":"ref38","first-page":"13937","article-title":"Dynamicvit: Efficient vision transformers with dynamic token sparsification","volume":"34","author":"Rao","year":"2021","journal-title":"Advances in neural information processing systems"},{"key":"ref39","author":"Shang","year":"2024","journal-title":"Llava-prumerge: Adaptive token reduction for efficient large multimodal models"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00851"},{"key":"ref41","author":"Sun","year":"2023","journal-title":"A simple and effective pruning approach for large language models"},{"key":"ref42","author":"Tang","year":"2024","journal-title":"Quest: Query-aware sparsity for efficient long-context 11 m inference"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00078"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01185"},{"key":"ref45","author":"Touvron","year":"2023","journal-title":"Llama 2: Open foundation and fine-tuned chat models"},{"key":"ref46","author":"Wang","year":"2024","journal-title":"Qwen2-vl: Enhancing vision-language model\u2019s perception of the world at any resolution"},{"key":"ref47","author":"Wang","year":"2024","journal-title":"Qwen2-vl: Enhancing vision-language model\u2019s perception of the world at any resolution"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00208"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-acl.802"},{"key":"ref50","author":"Xiao","year":"2024","journal-title":"Duoattention: Efficient long-context llm inference with retrieval and streaming heads"},{"key":"ref51","author":"Xiao","year":"2024","journal-title":"Efficient streaming language models with attention sinks"},{"key":"ref52","author":"Xing","year":"2024","journal-title":"Pyramiddrop: Accelerating your large visionlanguage models via pyramid visual redundancy reduction"},{"key":"ref53","author":"Xue","year":"2024","journal-title":"Longvila: Scaling long-context visual language models for long videos"},{"key":"ref54","author":"Yang","year":"2024","journal-title":"Qwen2 technical report"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.01843"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52734.2025.02777"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01054"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01100"},{"key":"ref59","author":"Zhang","year":"2024","journal-title":"Long context transfer from language to vision"},{"key":"ref60","author":"Zhang","year":"2024","journal-title":"Sparsevlm: Visual token sparsification for efficient vision-language model inference"},{"key":"ref61","author":"Zhang","year":"2024","journal-title":"Llavanext: A strong zero-shot video understanding model"},{"key":"ref62","author":"Zhang","year":"2023","journal-title":"H2O: Heavy-hitter oracle for efficient generative inference of large language models"},{"key":"ref63","author":"Zhou","year":"2024","journal-title":"Mlvu: A comprehensive benchmark for multi-task long video understanding"},{"key":"ref64","author":"Zhu","year":"2021","journal-title":"Vision transformer pruning"}],"event":{"name":"2025 IEEE\/CVF International Conference on Computer Vision (ICCV)","location":"Honolulu, HI, USA","start":{"date-parts":[[2025,10,19]]},"end":{"date-parts":[[2025,10,25]]}},"container-title":["2025 IEEE\/CVF International Conference on Computer Vision (ICCV)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11443115\/11443287\/11444431.pdf?arnumber=11444431","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T05:06:01Z","timestamp":1777611961000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11444431\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":64,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.02208","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}