通过分析视觉语言模型内部表示，减少幻觉并实现零样本分割。
Interpreting and Editing Vision-Language Representations to Mitigate Hallucinations
- 将图像特征投影到语言词汇，区分真实与幻觉物体。
- 在COCO2014上减少25.7%幻觉，同时保持模型性能。
- 可精准定位真实物体，适合需要高可靠性的应用。
我们研究视觉语言模型（VLMs）的内部表示以应对幻觉问题，尽管模型规模和训练不断进步。将VLM的内部图像表示投影至语言词汇后，发现对真实物体的输出概率高于幻觉物体。利用这些概率可空间定位真实物体。基于此，我们提出一种知识擦除算法，通过线性正交化图像特征与幻觉物体特征来消除幻觉。实验表明，对模型隐含表示进行定向编辑，可在COCO2014数据集上将幻觉降低25.7%，同时保持性能。研究揭示了深入理解VLM隐含表示对提升可靠性及实现新能力（如零样本分割）的重要作用。
原文摘要 · Abstract (English)
We investigate the internal representations of vision-language models (VLMs) to address hallucinations, a persistent challenge despite advances in model size and training. We project VLMs' internal image representations to their language vocabulary and observe more confident output probabilities on real objects than hallucinated objects. We additionally use these output probabilities to spatially localize real objects. Building on this approach, we introduce a knowledge erasure algorithm that removes hallucinations by linearly orthogonalizing image features with respect to hallucinated object features. We show that targeted edits to a model's latent representations can reduce hallucinations by up to 25.7% on the COCO2014 dataset while preserving performance. Our findings demonstrate how a deeper understanding of VLMs' latent representations can enhance reliability and enable novel capabilities, such as zero-shot segmentation.
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。