{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T03:29:19Z","timestamp":1777865359663,"version":"3.51.4"},"reference-count":42,"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\/iccv51701.2025.02221","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"23958-23967","source":"Crossref","is-referenced-by-count":0,"title":["AirCache: Activating Inter-Modal Relevancy KV Cache Compression for Efficient Large Vision-Language Model Inference"],"prefix":"10.1109","author":[{"given":"Kai","family":"Huang","sequence":"first","affiliation":[{"name":"Alibaba Group"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Hao","family":"Zou","sequence":"additional","affiliation":[{"name":"Alibaba Group"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Bochen","family":"Wang","sequence":"additional","affiliation":[{"name":"Alibaba Group"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ye","family":"Xi","sequence":"additional","affiliation":[{"name":"Alibaba Group"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zhen","family":"Xie","sequence":"additional","affiliation":[{"name":"Alibaba Group"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Hao","family":"Wang","sequence":"additional","affiliation":[{"name":"Alibaba Group"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Qwen-vl: A frontier large vision-language model with versatile abilities","author":"Bai","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00439"},{"key":"ref3","article-title":"Pyramidkv: Dynamic kv cache compression based on pyramidal information funneling","author":"Cai","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52733.2024.01311"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73004-7_2"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52733.2024.02283"},{"key":"ref7","article-title":"Mobilevlm: A fast, reproducible and strong vision language assistant for mobile devices","author":"Chu","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1145\/3664647.3685520"},{"key":"ref9","article-title":"Mmbench-video: A long-form multi-shot benchmark for holistic video understanding","author":"Fang","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref10","article-title":"Mme: A comprehensive evaluation benchmark for multimodal large language models","volume":"abs\/2306.13394","author":"Fu","year":"2023","journal-title":"ArXiv"},{"key":"ref11","article-title":"Zipvl: Efficient large vision-language models with dynamic token sparsification and kv cache compression","author":"He","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref12","article-title":"Matryoshka query transformer for large vision-language models","author":"Hu","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72643-9_13"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00271"},{"key":"ref15","article-title":"Compressed context memory for online language model interaction","author":"Kim","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref16","article-title":"Llava-onevision: Easy visual task transfer","author":"Li","year":"2024","journal-title":"arXiv preprint arXiv"},{"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"},{"key":"ref18","article-title":"Tokenpacker: Efficient visual projector for multimodal 11 m","author":"Li","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref19","article-title":"Snapkv: Llm knows what you are looking for before generation","author":"Li","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72952-2_19"},{"key":"ref21","first-page":"139997","article-title":"Minicache: Kv cache compression in depth dimension for large language models","volume":"37","author":"Liu","year":"2025","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02484"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1516"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72658-3_13"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72643-9_4"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.findings-acl.177"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/WACV51458.2022.00264"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/WACV48630.2021.00225"},{"key":"ref29","article-title":"Dynamic memory compression: Retrofitting l l ms for accelerated inference","author":"Nawrot","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref30","first-page":"606","article-title":"Efficiently scaling transformer inference","volume-title":"Proceedings of Machine Learning and Systems","volume":"5","author":"Pope","year":"2023"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00851"},{"key":"ref32","first-page":"87310","article-title":"Cambrian-1: A fully open, vision-centric exploration of multimodal 11 ms","volume":"37","author":"Tong","year":"2025","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref33","article-title":"Vl-cache: Sparsity and modality-aware kv cache compression for vision-language model inference acceleration","author":"Tu","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref34","article-title":"Prefixkv: Adaptive prefix kv cache is what vision instructionfollowing models need for efficient generation","author":"Wang","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref35","article-title":"Qwen2-vl: Enhancing vision-language model\u2019s perception of the world at any resolution","author":"Wang","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref36","article-title":"Deepseek-vl2: Mixture-ofexperts vision-language models for advanced multimodal understanding","author":"Wu","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref37","article-title":"Qwen2. 5 technical report","author":"Yang","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref38","article-title":"Kvsharer: Efficient inference via layer-wise dissimilar kv cache sharing","author":"Yang","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref39","article-title":"[cls] attention is all you need for trainingfree visual token pruning: Make vlm inference faster","author":"Zhang","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref40","article-title":"Cam: Cache merging for memory-efficient 11 ms inference","volume-title":"Forty-first International Conference on Machine Learning","author":"Zhang","year":"2024"},{"key":"ref41","article-title":"Video instruction tuning with synthetic data","author":"Zhang","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref42","article-title":"H2o: Heavy-hitter oracle for efficient generative inference of large language models","volume":"36","author":"Zhang","year":"2024","journal-title":"Advances in Neural Information Processing Systems"}],"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\/11445629.pdf?arnumber=11445629","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,30]],"date-time":"2026-04-30T06:20:28Z","timestamp":1777530028000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11445629\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":42,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.02221","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}