{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/silkie-preference-distillation-for-large","title":"Silkie: Preference Distillation for Large Visual Language Models","arxiv_id":"2312.10665","date":"2023-12-17","proceeding":null,"authors":["Lei LI","Zhihui Xie","Mukai Li","Shunian Chen","Peiyi Wang","Liang Chen","Yazheng Yang","Benyou Wang","Lingpeng Kong"],"abstract":"This paper explores preference distillation for large vision language models (LVLMs), improving their ability to generate helpful and faithful responses anchoring the visual context. We first build a vision-language feedback (VLFeedback) dataset utilizing AI annotation. Specifically, responses are generated by models sampled from 12 LVLMs, conditioned on multi-modal instructions sourced from various datasets. We adopt GPT-4V to assess the generated outputs regarding helpfulness, visual faithfulness, and ethical considerations. Furthermore, the preference supervision is distilled into Qwen-VL-Chat through the direct preference optimization (DPO) method. The resulting model Silkie, achieves 6.9% and 9.5% relative improvement on the MME benchmark regarding the perception and cognition capabilities, respectively. Silkie also demonstrates reduced hallucination by setting a new state-of-the-art score of 3.02 on the MMHal-Bench benchmark. Further analysis shows that DPO with our VLFeedback dataset mainly boosts the fine-grained perception and complex cognition abilities of LVLMs, leading to more comprehensive improvements compared to human-annotated preference datasets.","url_abs":"https://arxiv.org/abs/2312.10665v1","url_pdf":"https://arxiv.org/pdf/2312.10665v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[],"tasks":[{"task_slug":"hallucination","task_name":"Hallucination"},{"task_slug":"mme","task_name":"MME"},{"task_slug":"visual-question-answering-1","task_name":"Visual Question Answering"}],"methods":[{"method_slug":"dpo","method_name":"DPO"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/visual-question-answering-on-mm-vet","task":"Visual Question Answering","dataset":"MM-Vet","model":"Silkie","rank_in_archive_order":66,"of":231,"metrics":{"GPT-4 score":"49.9","Params":"7B"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2312.10665","atlas_url":"https://app.syntology.ai/?focus=2312.10665","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}