提出单次通过的无参考编辑方法，有效抑制视觉语言模型幻觉。
HulluEdit: Single-Pass Evidence-Consistent Subspace Editing for Mitigating Hallucinations in Large Vision-Language Models
- 通过正交子空间分解，分离视觉证据与错误先验
- 在多个基准上显著降低幻觉率，保持模型通用能力
- 无需参考模型，推理高效，适合实际部署
大型视觉语言模型（LVLMs）中的对象幻觉严重阻碍其可靠应用。现有方法难以兼顾效率与准确性：通常需要昂贵的参考模型和多次前向传播，或使用静态编辑会抑制真实视觉证据。为此，我们提出HulluEdit，一种单次通过、无需参考的干预框架。核心创新是正交子空间编辑：将模型隐藏状态分解为正交子空间——视觉证据、冲突先验和残余不确定性，实现对幻觉模式的选择性抑制，而不干扰视觉基础。该方法在数学上保证对先验子空间的编辑不会影响视觉成分。大量实验表明，HulluEdit在包括POPE和CHAIR在内的多个基准上达到当前最优幻觉减少效果，同时在MME上保持通用能力，并维持高效推理。方法持续优于对比解码和静态子空间编辑基线，为更可信的LVLM提供新路径。
原文摘要 · Abstract (English)
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.
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。