{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,11,1]],"date-time":"2026-11-01T03:57:54Z","timestamp":1793505474959,"version":"4.5.0"},"reference-count":44,"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\/iccv51701.2025.00348","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"3650-3661","source":"Crossref","is-referenced-by-count":4,"title":["Large Multi-modal Models Can Interpret Features in Large Multi-modal Models"],"prefix":"10.1109","author":[{"given":"Kaichen","family":"Zhang","sequence":"first","affiliation":[{"name":"S-Lab, NTU,Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yifei","family":"Shen","sequence":"additional","affiliation":[{"name":"LMMs-Lab Team"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bo","family":"Li","sequence":"additional","affiliation":[{"name":"S-Lab, NTU,Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ziwei","family":"Liu","sequence":"additional","affiliation":[{"name":"S-Lab, NTU,Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1137\/140979861"},{"key":"ref2","volume-title":"Apple. Apple intelligence is available today on iphone, ipad and mac","year":"2024"},{"key":"ref3","article-title":"Qwen-vl: A versatile vision-language model for understanding, localization, text reading, and beyond","author":"Bai","year":"2023","journal-title":"arXiv preprint"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.354"},{"key":"ref5","volume-title":"Language models can explain neurons in language models","author":"Bills","year":"2023"},{"key":"ref6","article-title":"Towards monosemanticity: Decomposing language models with dictionary learning","author":"Bricken","year":"2023","journal-title":"Transformer Circuits Thread"},{"key":"ref7","author":"Chen","year":"2024","journal-title":"Red teaming gpt-4v: Are gpt-4v safe against uni\/multi-modal jailbreak attacks?"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52733.2024.02283"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2142"},{"key":"ref10","author":"Dubey","year":"2024","journal-title":"The llama 3 herd of models"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1007\/978-1-4419-7011-4"},{"key":"ref12","article-title":"Toy models of superposition","author":"Elhage","year":"2022","journal-title":"arXiv preprint"},{"key":"ref13","author":"Fu","year":"2024","journal-title":"Mme: A comprehensive evaluation benchmark for multimodal large language models"},{"key":"ref14","author":"Gao","year":"2024","journal-title":"Scaling and evaluating sparse autoencoders"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52733.2024.01363"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00371"},{"key":"ref17","author":"Li","year":"2024","journal-title":"Llava-next: Stronger llms supercharge multimodal capabilities in the wild"},{"key":"ref18","author":"Li","year":"2024","journal-title":"Llava-onevision: Easy visual task transfer"},{"key":"ref19","author":"Li","year":"2023","journal-title":"Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.blackboxnlp-1.19"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-10602-1_48"},{"key":"ref22","author":"Liu","year":"2024","journal-title":"Llava-next: Improved reasoning, ocr, and world knowledge"},{"key":"ref23","article-title":"Visual instruction tuning","volume":"36","author":"Liu","year":"2024","journal-title":"Advances in neural information processing systems"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72970-6_3"},{"key":"ref25","article-title":"Reducing hallucinations in vision-language models via latent space steering","author":"Liu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref26","article-title":"A survey on vision-language-action models for embodied ai","author":"Ma","year":"2024","journal-title":"arXiv preprint"},{"key":"ref27","author":"Makhzani","year":"2014","journal-title":"k-sparse autoencoders"},{"key":"ref28","volume-title":"Meta. Introducing orion: Our first true augmented reality glasses","year":"2024"},{"key":"ref29","author":"Nanda","year":"2023","journal-title":"Attribution patching: Activation patching at industrial scale"},{"key":"ref30","author":"Neo","year":"2024","journal-title":"Towards interpreting visual information processing in vision-language models"},{"key":"ref31","doi-asserted-by":"crossref","DOI":"10.23915\/distill.00007","volume-title":"Feature visualization","author":"Olah","year":"2017"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.23915\/distill.00024.001"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1038\/381607a0"},{"key":"ref34","author":"Parekh","year":"2024","journal-title":"A concept-based explainability framework for large multimodal models"},{"key":"ref35","author":"Schaeffer","year":"2024","journal-title":"When do universal image jailbreaks transfer between vision-language models"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00158"},{"key":"ref37","article-title":"Scaling monosemanticity: Extracting interpretable features from claude 3 sonnet","author":"Templeton","year":"2024","journal-title":"Transformer Circuits Thread"},{"key":"ref38","author":"Tong","year":"2024","journal-title":"Cambrian-1: A fully open, vision-centric exploration of multimodal 11 ms"},{"key":"ref39","author":"Wang","year":"2022","journal-title":"Interpretability in the wild: a circuit for indirect object identification in gpt-2 small"},{"key":"ref40","author":"Wang","year":"2024","journal-title":"Qwen2-vl: Enhancing vision-language model\u2019s perception of the world at any resolution"},{"issue":"1","key":"ref41","first-page":"1","article-title":"The dawn of lmms: Preliminary explorations with gpt-4v (ision)","volume":"9","author":"Yang","year":"2023","journal-title":"arXiv preprint"},{"key":"ref42","author":"Yu","year":"2023","journal-title":"Mm -vet: Evaluating large multimodal models for integrated capabilities"},{"key":"ref43","author":"Zheng","year":"2024","journal-title":"Sglang: Efficient execution of structured language model programs"},{"key":"ref44","article-title":"Object detectors emerge in deep scene cnns","author":"Zhou","year":"2014","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\/11444372.pdf?arnumber=11444372","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,30]],"date-time":"2026-04-30T06:36:50Z","timestamp":1777531010000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11444372\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":44,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.00348","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}