HulluEdit: 单次通过证据一致子空间编辑用于缓解大视觉-语言模型中的幻觉
HulluEdit: Single-Pass Evidence-Consistent Subspace Editing for Mitigating Hallucinations in Large Vision-Language Models
AI总结：
HulluEdit通过单次通过的正交子空间编辑方法，有效缓解大视觉-语言模型中的幻觉问题，实现了最先进的幻觉减少效果。
摘要：
在大型视觉-语言模型（LVLMs）中，对象幻觉显著阻碍了其可靠部署。现有方法难以在效率和准确性之间取得平衡：它们通常需要昂贵的参考模型和多次前向传递，或者应用静态编辑，这可能会抑制真实的视觉证据。为了解决这个问题，我们引入了HulluEdit，一种单次通过、无参考干预框架。我们的核心创新是正交子空间编辑：我们将模型的隐藏状态分解为正交子空间——视觉证据、冲突先验和残差不确定性——从而能够选择性地抑制幻觉模式而不影响视觉基础。这种方法在数学上保证了对先验子空间的编辑不会影响视觉部分。广泛的实验表明，HulluEdit在POPE和CHAIR等基准上实现了最先进的幻觉减少，适用于各种架构，同时在MME上保持一般能力并在推理中保持高效。我们的方法在对比解码和静态子空间编辑基线中始终表现更优，为更可信的LVLMs提供了一条新路径。
英文摘要：
Object hallucination in Large Vision-Language Models (LVLMs) significantly hinders their reliable deployment. Existing methods struggle to balance efficiency and accuracy: they often require expensive reference models and multiple forward passes, or apply static edits that risk suppressing genuine visual evidence. To address this, we introduce HulluEdit, a single-pass, reference-free intervention framework. Our core innovation is orthogonal subspace editing: we decompose the hidden states of the model into orthogonal subspaces - visual evidence, conflicting priors, and residual uncertainty - enabling selective suppression of hallucinatory patterns without interfering with visual grounding. This approach mathematically guarantees that edits applied to the prior subspace leave the visual component entirely unaffected. Extensive experiments show that HulluEdit achieves state-of-the-art hallucination reduction on benchmarks including POPE and CHAIR across diverse architectures, while preserving general capabilities on MME and maintaining efficient inference. Our method consistently outperforms contrastive decoding and static subspace editing baselines, offering a new pathway toward more trustworthy LVLMs.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
