{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T10:22:21Z","timestamp":1777890141517,"version":"3.51.4"},"reference-count":45,"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"}],"funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"publisher","award":["62325405,62104128,U19B2019,U21B2031,61832007,62204164,92364201"],"award-info":[{"award-number":["62325405,62104128,U19B2019,U21B2031,61832007,62204164,92364201"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,10,19]]},"DOI":"10.1109\/iccv51701.2025.02103","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"22654-22663","source":"Crossref","is-referenced-by-count":0,"title":["FrameFusion: Combining Similarity and Importance for Video Token Reduction on Large Vision Language Models"],"prefix":"10.1109","author":[{"given":"Tianyu","family":"Fu","sequence":"first","affiliation":[{"name":"Tsinghua University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Tengxuan","family":"Liu","sequence":"additional","affiliation":[{"name":"Tsinghua University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Qinghao","family":"Han","sequence":"additional","affiliation":[{"name":"Peking University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Guohao","family":"Dai","sequence":"additional","affiliation":[{"name":"Shanghai Jiao Tong University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Shengen","family":"Yan","sequence":"additional","affiliation":[{"name":"Infinigence-AI"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Huazhong","family":"Yang","sequence":"additional","affiliation":[{"name":"Tsinghua University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Xuefei","family":"Ning","sequence":"additional","affiliation":[{"name":"Tsinghua University"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yu","family":"Wang","sequence":"additional","affiliation":[{"name":"Tsinghua University"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","author":"Cai","year":"2024","journal-title":"Pyramidkv: Dynamic kv cache compression based on pyramidal information funneling"},{"key":"ref2","author":"Chai","year":"2024","journal-title":"Auroracap: Efficient, performant video detailed captioning and a new benchmark"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73004-7_2"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00205"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52733.2024.02283"},{"key":"ref6","author":"Dong","year":"2024","journal-title":"Internlm-xcomposer24khd: A pioneering large vision-language model handling resolutions from 336 pixels to 4 k hd"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.02245"},{"key":"ref8","author":"Fu","year":"2024","journal-title":"Lazyllm: Dynamic token pruning for efficient long context 11 m inference"},{"key":"ref9","author":"Fu","year":"2024","journal-title":"Moa: Mixture of sparse attention for automatic large language model compression"},{"key":"ref10","volume":"abs\/2310.01801","author":"Ge","year":"2023","journal-title":"Model tells you what to discard: Adaptive kv cache compression for 11 ms"},{"key":"ref11","author":"Han","year":"2023","journal-title":"Lm-infinite: Simple on-the-fly length generalization for large language models"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.01282"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.19"},{"key":"ref14","author":"Jiang","year":"2024","journal-title":"Minference 1.0: Accelerating pre-filling for long-context 11 ms via dynamic sparse attention"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1007\/s44267-025-00099-6"},{"key":"ref16","author":"Li","year":"2024","journal-title":"Llava-onevision: Easy visual task transfer"},{"key":"ref17","first-page":"19730","article-title":"Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models","volume-title":"International conference on machine learning","author":"Li","year":"2023"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72698-9_17"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02095"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0722"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72952-2_19"},{"key":"ref22","volume":"abs\/2305.17118","author":"Liu","year":"2023","journal-title":"Scissorhands: Exploiting the persistence of importance hypothesis for 11 m kv cache compression at test time"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.00390"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2004"},{"key":"ref25","author":"Shang","year":"2024","journal-title":"Llava-prumerge: Adaptive token reduction for efficient large multimodal models"},{"key":"ref26","author":"Tang","year":"2024","journal-title":"Quest: Query-aware sparsity for efficient long-context 11 m inference"},{"key":"ref27","author":"Team","year":"2024","journal-title":"Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context"},{"key":"ref28","author":"Tu","year":"2024","journal-title":"Vl-cache: Sparsity and modality-aware kv cache compression for vision-language model inference acceleration"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA51647.2021.00018"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73414-4_26"},{"key":"ref31","article-title":"Efficient streaming language models with attention sinks","volume-title":"The Twelfth International Conference on Learning Representations","author":"Xiao","year":"2024"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00965"},{"key":"ref33","author":"Xu","year":"2024","journal-title":"Slowfast-llava: A strong training-free baseline for video large language models"},{"key":"ref34","author":"Yao","year":"2024","journal-title":"Minicpm-v: A gpt-4v level mllm on your phone"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1002\/0470011815.b2a15150"},{"key":"ref36","volume-title":"Sageattention2: Efficient attention with thorough outlier smoothing and per-thread int4 quantization","author":"Zhang","year":"2024"},{"key":"ref37","author":"Zhang","year":"2024","journal-title":"A-vl: Adaptive attention for large visionlanguage models"},{"key":"ref38","article-title":"Sageattention: Accurate 8-bit attention for plug-andplay inference acceleration","volume-title":"International Conference on Learning Representations (ICLR)","author":"Zhang","year":"2025"},{"key":"ref39","volume-title":"Spargeattn: Accurate sparse attention accelerating any model inference","author":"Zhang","year":"2025"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-naacl.51"},{"key":"ref41","author":"Zhang","year":"2024","journal-title":"Sparsevlm: Visual token sparsification for efficient vision-language model inference"},{"key":"ref42","author":"Zhang","year":"2024","journal-title":"Video instruction tuning with synthetic data"},{"key":"ref43","volume":"abs\/2306.14048","author":"Zhang","year":"2023","journal-title":"H2o: Heavy-hitter oracle for efficient generative inference of large language models"},{"key":"ref44","author":"Zhao","year":"2024","journal-title":"Needle in a video haystack: A scalable synthetic framework for benchmarking video mllms"},{"key":"ref45","author":"Zhong","year":"2024","journal-title":"Aim: Adaptive inference of multi-modal 11 ms via token merging and 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\/11444404.pdf?arnumber=11444404","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T05:29:34Z","timestamp":1777613374000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11444404\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":45,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.02103","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}