用视觉引导强化学习，减少大模型幻觉问题。
V-DPO: Mitigating Hallucination in Large Vision Language Models via Vision-Guided Direct Preference Optimization
- 通过视觉引导的直接偏好优化，降低语言模型对视觉输入的依赖偏差。
- 在多种幻觉评测中显著优于基线方法，尤其在图像对比数据上表现突出。
- 适合关注多模态对齐与生成准确性的研究人员和开发者。
大型视觉语言模型（LVLM）存在幻觉问题，导致文本输出与输入视觉内容不一致。研究发现，过度依赖语言模型主干是原因之一，会引入语言先验偏见，削弱对视觉信息的注意力。为此，我们提出视觉引导的直接偏好优化（V-DPO），在训练阶段增强视觉上下文学习能力。为评估其在不同训练数据上的有效性，我们构建了一个包含响应-图像对比偏好对的合成数据集，并与现有真人标注的幻觉样本进行对比。实验表明，相较于基线方法，本方法在多个幻觉评测基准上均取得显著提升。分析显示，V-DPO在图像对比偏好数据上表现尤为优异，展现出更强的视觉细节理解和捕捉能力。代码已公开于 https://github.com/YuxiXie/V-DPO。
原文摘要 · Abstract (English)
Large vision-language models (LVLMs) suffer from hallucination, resulting in misalignment between the output textual response and the input visual content. Recent research indicates that the over-reliance on the Large Language Model (LLM) backbone, as one cause of the LVLM hallucination, inherently introduces bias from language priors, leading to insufficient context attention to the visual inputs. We tackle this issue of hallucination by mitigating such over-reliance through preference learning. We propose Vision-guided Direct Preference Optimization (V-DPO) to enhance visual context learning at training time. To interpret the effectiveness and generalizability of V-DPO on different types of training data, we construct a synthetic dataset containing both response- and image-contrast preference pairs, compared against existing human-annotated hallucination samples. Our approach achieves significant improvements compared with baseline methods across various hallucination benchmarks. Our analysis indicates that V-DPO excels in learning from image-contrast preference data, demonstrating its superior ability to elicit and understand nuances of visual context. Our code is publicly available at https://github.com/YuxiXie/V-DPO.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。