通过电路看见：面向视觉变换器的忠实机制可解释性
Seeing Through Circuits: Faithful Mechanistic Interpretability for Vision Transformers
AI总结：
本文研究了通过计算图在视觉变换器中识别有用机制电路的可能性，提出自动视觉电路发现方法，发现分类特定电路、CLIP中的文字攻击电路以及可引导纠正有害行为的电路。
摘要：
神经网络内部推理的透明性是可解释性研究的核心，有助于信任、安全和理解这些模型。机制可解释性领域近期专注于研究任务特定的计算图，由模型组件之间的连接（边）定义。此类基于边的电路已在大语言模型中定义，但基于视觉的方法迄今为止只考虑基于神经元的电路。这些电路只说明了哪些信息被编码，而没有说明信息如何通过复杂的布线在神经网络中路由。本文研究了是否可以通过计算图在视觉变换器中识别有用的机制电路。我们提出了一种有效的自动视觉电路发现（Vi-CD）方法，该方法恢复分类特定的电路，识别CLIP中基于文字攻击的电路，并发现可引导纠正有害模型行为的电路。总体而言，我们发现可以从视觉变换器中恢复具有洞察力和可操作性的基于边的电路，从而增加这些模型内部计算的透明性。
英文摘要：
Transparency of neural networks' internal reasoning is at the heart of interpretability research, adding to trust, safety, and understanding of these models. The field of mechanistic interpretability has recently focused on studying task-specific computational graphs, defined by connections (edges) between model components. Such edge-based circuits have been defined in the context of large language models, yet vision-based approaches so far only consider neuron-based circuits. These tell which information is encoded, but not how it is routed through the complex wiring of a neural network. In this work, we investigate whether useful mechanistic circuits can be identified through computational graphs in vision transformers. We propose an effective method for Automatic Visual Circuit Discovery (Vi-CD) that recovers class-specific circuits for classification, identifies circuits underlying typographic attacks in CLIP, and discovers circuits that lend themselves for steering to correct harmful model behavior. Overall, we find that insightful and actionable edge-based circuits can be recovered from vision transformers, adding transparency to the internal computations of these models.
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。
