视觉语言模型有坏眼睛吗？通过机制可解释性诊断组合失败
Do VLMs Have Bad Eyes? Diagnosing Compositional Failures via Mechanistic Interpretability
AI总结：
本文通过机制可解释性方法诊断视觉语言模型组合失败的原因，发现CLIP视觉编码器MLP层中神经元的特征叠加现象直接损害了组合特征表示与对象绑定能力。
摘要：
视觉语言模型（VLMs）在整合视觉与文本信息以完成图像描述和视觉问答等任务方面表现出色。然而，这些模型在组合泛化和对象绑定方面存在困难，这限制了它们处理对象及其属性的新颖组合的能力。我们的工作利用机制可解释性技术探索这些失败的根本原因。我们展示证据表明，CLIP视觉编码器的MLP层中的单个神经元表示多个特征，而这种“叠加”直接阻碍了其组合特征表示，进而影响组合推理和对象绑定能力。我们希望这项研究能成为揭示视觉语言模型中组合失败的机制根源的第一步。代码和支持结果可在 https://github.com/Mystic-Slice/Do-VLMs-Have-Bad-Eyes 找到。
英文摘要：
Vision-Language Models (VLMs) have shown remarkable performance in integrating visual and textual information for tasks such as image captioning and visual question answering. However, these models struggle with compositional generalization and object binding, which limit their ability to handle novel combinations of objects and their attributes. Our work explores the root causes of these failures using mechanistic interpretability techniques. We show evidence that individual neurons in the MLP layers of CLIP's vision encoder represent multiple features, and this "superposition" directly hinders its compositional feature representation which consequently affects compositional reasoning and object binding capabilities. We hope this study will serve as an initial step toward uncovering the mechanistic roots of compositional failures in VLMs. The code and supporting results can be found https://github.com/Mystic-Slice/Do-VLMs-Have-Bad-Eyes.
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。
