{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T10:12:38Z","timestamp":1777889558976,"version":"3.51.4"},"reference-count":56,"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\/100019923","name":"DEVCOM Army Research Laboratory","doi-asserted-by":"publisher","award":["W911QX-21-D-0001"],"award-info":[{"award-number":["W911QX-21-D-0001"]}],"id":[{"id":"10.13039\/100019923","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\/iccv51701.2025.00077","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"741-751","source":"Crossref","is-referenced-by-count":0,"title":["Structured Policy Optimization: Enhance Large Vision-Language Model via Self-Referenced Dialogue"],"prefix":"10.1109","author":[{"given":"Guohao","family":"Sun","sequence":"first","affiliation":[{"name":"Rochester Institute of Technology"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Can","family":"Qin","sequence":"additional","affiliation":[{"name":"Salesforce AI Research"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yihao","family":"Feng","sequence":"additional","affiliation":[{"name":"Salesforce AI Research"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zeyuan","family":"Chen","sequence":"additional","affiliation":[{"name":"Salesforce AI Research"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ran","family":"Xu","sequence":"additional","affiliation":[{"name":"Salesforce AI Research"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Sohail","family":"Dianat","sequence":"additional","affiliation":[{"name":"Rochester Institute of Technology"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Majid","family":"Rabbani","sequence":"additional","affiliation":[{"name":"Rochester Institute of Technology"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Raghuveer","family":"Rao","sequence":"additional","affiliation":[{"name":"DEVCOM Army Research Laboratory"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zhiqiang","family":"Tao","sequence":"additional","affiliation":[{"name":"Rochester Institute of Technology"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","volume-title":"Gpt-4 technical report","author":"Achiam","year":"2023"},{"key":"ref2","article-title":"Training a helpful and harmless assistant with reinforcement learning from human feedback","author":"Bai","year":"2022","journal-title":"arXiv preprint arXiv"},{"key":"ref3","article-title":"Comparing bad apples to good oranges: Aligning large language models via joint preference optimization","author":"Bansal","year":"2024","journal-title":"ArXiv"},{"key":"ref4","article-title":"Sharegpt4v: Improving large multi-modal models with better captions","author":"Chen","year":"2023","journal-title":"ArXiv"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0850"},{"key":"ref6","article-title":"Self-play fine-tuning converts weak language models to strong language models","author":"Chen","year":"2024","journal-title":"ArXiv"},{"key":"ref7","article-title":"Internvl: Scaling up vision foundation models and aligning for generic visual-linguistic tasks","volume-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition","author":"Chen","year":"2024"},{"key":"ref8","article-title":"Deep reinforcement learning from human preferences","volume-title":"Proceedings of the 31st International Conference on Neural Information Processing Systems","year":"2017"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1016\/j.tate.2023.104068"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2142"},{"key":"ref11","article-title":"Enhancing large vision language models with self-training on image comprehension","author":"Deng","year":"2024","journal-title":"NeurIPS"},{"key":"ref12","article-title":"Raft: Reward ranked finetuning for generative foundation model alignment","author":"Dong","year":"2023","journal-title":"ArXiv"},{"key":"ref13","article-title":"Mme: A comprehensive evaluation benchmark for multimodal large language models","author":"Fu","year":"2023","journal-title":"ArXiv"},{"key":"ref14","article-title":"Video-mme: The first-ever comprehensive evaluation benchmark of multi-modal 11 ms in video analysis","author":"Fu","year":"2024","journal-title":"ArXiv"},{"key":"ref15","article-title":"Lora: Low-rank adaptation of large language models","author":"Hu","year":"2022","journal-title":"ICLR"},{"key":"ref16","article-title":"Mantis: Interleaved multiimage instruction tuning","author":"Jiang","year":"2024","journal-title":"ArXiv"},{"key":"ref17","article-title":"Step-dpo: Step-wise preference optimization for long-chain reasoning of 11 ms","author":"Lai","year":"2024","journal-title":"ArXiv"},{"key":"ref18","article-title":"Llava-onevision: Easy visual task transfer","author":"Li","year":"2024","journal-title":"ArXiv"},{"key":"ref19","article-title":"Llava-med: Training a large language-and-vision assistant for biomedicine in one day","author":"Li","year":"2023","journal-title":"NeurIPS"},{"key":"ref20","article-title":"Llava-next-interleave: Tackling multi-image, video, and 3d in large multimodal models","author":"Li","year":"2024","journal-title":"ArXiv"},{"key":"ref21","article-title":"Fine-tuning multimodal 11 ms to follow zero-shot demonstrative instructions","volume-title":"International Conference on Learning Representations","author":"Li","year":"2024"},{"key":"ref22","article-title":"Silkie: Preference distillation for large visual language models","author":"Li","year":"2023","journal-title":"ArXiv"},{"key":"ref23","article-title":"Llama-vid: An image is worth 2 tokens in large language models","author":"Li","year":"2023","journal-title":"ArXiv"},{"key":"ref24","article-title":"Improved baselines with visual instruction tuning","author":"Liu","year":"2023","journal-title":"ArXiv"},{"key":"ref25","article-title":"Egoschema: A diagnostic benchmark for very longform video language understanding","author":"Mangalam","year":"2023","journal-title":"ArXiv"},{"key":"ref26","article-title":"Gpt-4v(ision) system card","journal-title":"2023"},{"key":"ref27","article-title":"OpenAI","year":"2024","journal-title":"Gpt-4o system card"},{"key":"ref28","article-title":"Training language models to follow instructions with human feedback","author":"Ouyang","year":"2022","journal-title":"NeurIPS"},{"key":"ref29","article-title":"Direct preference optimization: Your language model is secretly a reward model","author":"Rafailov","year":"2023","journal-title":"NeurIPS"},{"key":"ref30","article-title":"From r to q: Your language model is secretly a q -function","volume-title":"First Conference on Language Modeling","author":"Rafailov","year":"2024"},{"key":"ref31","article-title":"Direct nash optimization: Teaching language models to selfimprove with general preferences","author":"Rosset","year":"2024","journal-title":"ArXiv"},{"key":"ref32","article-title":"Beyond human data: Scaling self-training for problem-solving with language models","author":"Singh","year":"2023","journal-title":"Trans. Mach. Learn. Res."},{"key":"ref33","article-title":"Preference ranking optimization for human alignment","author":"Song","year":"2023","journal-title":"ArXiv"},{"key":"ref34","article-title":"Stllava-med: Self-training large language and vision assistant for medical","author":"Sun","year":"2024","journal-title":"EMNLP"},{"key":"ref35","article-title":"Sq-llava: Self-questioning for large vision-language assistant","author":"Sun","year":"2024","journal-title":"ECCV"},{"key":"ref36","article-title":"Gemini: a family of highly capable multimodal models","author":"Team","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref37","article-title":"Muirbench: A comprehensive benchmark for robust multi-image understanding","author":"Wang","year":"2024","journal-title":"ArXiv"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.460"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73414-4_26"},{"key":"ref40","article-title":"Longvideobench: A benchmark for long-context interleaved video-language understanding","author":"Wu","year":"2024","journal-title":"ArXiv"},{"key":"ref41","article-title":"\u03b2 dpo: Direct preference optimization with dynamic \u03b2","author":"Wu","year":"2024","journal-title":"ArXiv"},{"key":"ref42","article-title":"Self-play preference optimization for language model alignment","author":"Wu","year":"2024","journal-title":"ArXiv"},{"key":"ref43","article-title":"Grok-1.5 vision preview","year":"2024","journal-title":"Technical report, x.ai"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00965"},{"key":"ref45","article-title":"Llava-critic: Learning to evaluate multimodal models","author":"Xiong","year":"2024","journal-title":"ArXiv"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1057\/s41599-023-01508-1"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52734.2025.02777"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00913"},{"key":"ref49","article-title":"Lmmseval: Reality check on the evaluation of large multimodal models","author":"Zhang","year":"2024","journal-title":"ArXiv"},{"key":"ref50","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"},{"key":"ref51","article-title":"Long context transfer from language to vision","author":"Zhang","year":"2024","journal-title":"ArXiv"},{"key":"ref52","article-title":"Llavar: Enhanced visual instruction tuning for text-rich image understanding","author":"Zhang","year":"2023","journal-title":"ArXiv"},{"key":"ref53","article-title":"Video instruction tuning with synthetic data","author":"Zhang","year":"2024","journal-title":"ArXiv"},{"key":"ref54","article-title":"Mlvu: A comprehensive benchmark for multi-task long video understanding","author":"Zhou","year":"2024","journal-title":"ArXiv"},{"key":"ref55","article-title":"Minigpt-4: Enhancing vision-language understanding with advanced large language models","author":"Zhu","year":"2023","journal-title":"ArXiv"},{"key":"ref56","article-title":"Fine-tuning language models from human preferences","author":"Ziegler","year":"2019","journal-title":"ArXiv"}],"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\/11444142.pdf?arnumber=11444142","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T05:12:57Z","timestamp":1777612377000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11444142\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":56,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.00077","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}