{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T10:22:46Z","timestamp":1777890166040,"version":"3.51.4"},"reference-count":60,"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\/iccv51701.2025.00353","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"3705-3715","source":"Crossref","is-referenced-by-count":0,"title":["p-MoD: Building Mixture-of-Depths MLLMs via Progressive Ratio Decay"],"prefix":"10.1109","author":[{"given":"Jun","family":"Zhang","sequence":"first","affiliation":[{"name":"Nanjing University,State Key Laboratory for Novel Software Technology"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Desen","family":"Meng","sequence":"additional","affiliation":[{"name":"Nanjing University,State Key Laboratory for Novel Software Technology"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zhengming","family":"Zhang","sequence":"additional","affiliation":[{"name":"China Mobile Research Institute"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zhenpeng","family":"Huang","sequence":"additional","affiliation":[{"name":"Nanjing University,State Key Laboratory for Novel Software Technology"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Tao","family":"Wu","sequence":"additional","affiliation":[{"name":"Nanjing University,State Key Laboratory for Novel Software Technology"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Limin","family":"Wang","sequence":"additional","affiliation":[{"name":"Nanjing University,State Key Laboratory for Novel Software Technology"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Pixtral 12b","author":"Agrawal","year":"2024","journal-title":"arXiv preprint"},{"key":"ref2","article-title":"Qwen-vl: A frontier large vision-language model with versatile abilities","author":"Bai","year":"2023","journal-title":"arXiv preprint"},{"key":"ref3","article-title":"Auroracap: Efficient, performant video detailed captioning and a new benchmark","author":"Chai","year":"2024","journal-title":"arXiv preprint"},{"key":"ref4","article-title":"Llavolta: Efficient multi-modal models via stage-wise visual context compression","author":"Chen","year":"2024","journal-title":"arXiv preprint"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73004-7_2"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1007\/s11432-024-4231-5"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.70"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2142"},{"issue":"120","key":"ref9","first-page":"1","article-title":"Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity","volume":"23","author":"Fedus","year":"2022","journal-title":"Journal of Machine Learning Research"},{"key":"ref10","article-title":"Mme: A comprehensive evaluation benchmark for multi-modal large language models","author":"Fu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.670"},{"key":"ref12","article-title":"Free video-llm: Prompt-guided visual perception for efficient training-free video 11 ms","author":"Han","year":"2024","journal-title":"arXiv preprint"},{"key":"ref13","article-title":"Zipvl: Efficient large vision-language models with dynamic token sparsification and kv cache compression","author":"He","year":"2024","journal-title":"arXiv preprint"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.70675\/802872a7z8092z4556za1fdz2c7af45f0e72"},{"key":"ref15","article-title":"illava: An image is worth fewer than 1\/3 input tokens in large multi-modal models","author":"Hu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref16","article-title":"Matryoshka query transformer for large vision-language models","author":"Hu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref17","article-title":"Mini-monkey: Alleviating the semantic sawtooth effect for lightweight mllms via complementary image pyramid","author":"Huang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00686"},{"key":"ref19","article-title":"Mixtral of experts","author":"Albert","year":"2024","journal-title":"arXiv preprint"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.3115\/v1\/D14-1086"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46493-0_15"},{"key":"ref22","article-title":"Gshard: Scaling giant models with conditional computation and automatic sharding","author":"Lepikhin","journal-title":"arXiv preprint"},{"key":"ref23","article-title":"Seed-bench: Benchmarking multi-modal 11 ms with generative comprehension","author":"Li","year":"2023","journal-title":"arXiv preprint"},{"key":"ref24","article-title":"Llava-onevision: Easy visual task transfer","author":"Li","year":"2024","journal-title":"arXiv preprint"},{"key":"ref25","article-title":"Videochat-flash: Hierarchical compression for long-context video modeling","author":"Li","year":"2024","journal-title":"arXiv preprint"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.20"},{"key":"ref27","author":"Liu","year":"2023","journal-title":"Improved baselines with visual instruction tuning"},{"key":"ref28","author":"Liu","year":"2024","journal-title":"Llava-next: Improved reasoning, ocr, and world knowledge"},{"key":"ref29","article-title":"Visual instruction tuning","volume":"36","author":"Liu","year":"2024","journal-title":"Advances in neural information processing systems"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72658-3_13"},{"key":"ref31","article-title":"Oryx mllm: On-demand spatial-temporal understanding at arbitrary resolution","author":"Liu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.52202\/068431-0182"},{"key":"ref33","article-title":"\u03b3-mod: Exploring mixture-of-depth adaptation for multi-modal large language models","author":"Luo","year":"2024","journal-title":"arXiv preprint"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00331"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.findings-acl.177"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/WACV48630.2021.00225"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/WACV51458.2022.00264"},{"key":"ref38","article-title":"Mm1: Methods, analysis & insights from multi-modal 11 m pre-training","author":"McKinzie","year":"2024","journal-title":"arXiv preprint"},{"key":"ref39","article-title":"Token sequence compression for efficient multi-modal computing","author":"Omri","year":"2025","journal-title":"arXiv preprint"},{"key":"ref40","first-page":"8748","article-title":"Learning transferable visual models from natural language supervision","volume-title":"International conference on machine learning","author":"Radford","year":"2021"},{"key":"ref41","article-title":"Mixture-of-depths: Dynamically allocating compute in transformer-based language models","author":"Raposo","year":"2024","journal-title":"arXiv preprint"},{"key":"ref42","article-title":"Llava-prumerge: Adaptive token reduction for efficient large multi-modal models","author":"Shang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref43","article-title":"Outrageously large neural networks: The sparsely-gated mixture-of-experts layer","author":"Shazeer","year":"2017","journal-title":"arXiv preprint"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00851"},{"key":"ref45","article-title":"Gemini: a family of highly capable multi-modal models","author":"Team","year":"2023","journal-title":"arXiv preprint"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2771"},{"key":"ref47","article-title":"Qwen2-vl: Enhancing vision-language model\u2019s perception of the world at any resolution","author":"Wang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref48","article-title":"Pyramiddrop: Accelerating your large vision-language models via pyramid visual redundancy reduction","author":"Xing","year":"2024","journal-title":"arXiv preprint"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52734.2025.01843"},{"key":"ref50","article-title":"Deco: Decoupling token compression from semantic abstraction in multi-modal large language models","author":"Yao","year":"2024","journal-title":"arXiv preprint"},{"key":"ref51","article-title":"mplug-owl3: Towards long image-sequence understanding in multi-modal large language models","author":"Ye","year":"2024","journal-title":"arXiv preprint"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46475-6_5"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00913"},{"key":"ref54","article-title":"Lmms-eval: Reality check on the evaluation of large multi-modal models","author":"Zhang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref55","article-title":"Internlm-xcomposer-2.5: A versatile large vision language model supporting long-contextual input and output","author":"Zhang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref56","article-title":"[cls] attention is all you need for training-free visual token pruning: Make vlm inference faster","author":"Zhang","year":"2024","journal-title":"arXiv eprints"},{"key":"ref57","article-title":"Sparsevlm: Visual token sparsification for efficient vision-language model inference","author":"Zhang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref58","article-title":"Beyond llava-hd: Diving into high-resolution large multi-modal models","author":"Zhang","year":"2024","journal-title":"arXiv preprint"},{"key":"ref59","article-title":"Minigpt-4: Enhancing vision-language understanding with advanced large language models","author":"Zhu","year":"2023","journal-title":"arXiv preprint"},{"key":"ref60","article-title":"St-moe: Designing stable and transferable sparse expert models","author":"Zoph","year":"2022","journal-title":"arXiv preprint"}],"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\/11445658.pdf?arnumber=11445658","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T05:32:25Z","timestamp":1777613545000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11445658\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":60,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.00353","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}