Many recent language model (LM) interpretability studies have adopted the circuits framework, which aims to find the minimal computational subgraph, or circuit, that explains LM behavior on a given task. Most studies determine which edges belong in a LM's circuit by performing causal interventions on each edge independently, but this scales poorly with model size. Edge attribution patching (EAP), gradient-based approximation to interventions, has emerged as a scalable but imperfect solution to this problem. In this paper, we introduce a new method - EAP with integrated gradients (EAP-IG) - that aims to better maintain a core property of circuits: faithfulness. A circuit is faithful if all model edges outside the circuit can be ablated without changing the model's performance on the task; faithfulness is what justifies studying circuits, rather than the full model. Our experiments demonstrate that circuits found using EAP are less faithful than those found using EAP-IG, even though both have high node overlap with circuits found previously using causal interventions. We conclude more generally that when using circuits to compare the mechanisms models use to solve tasks, faithfulness, not overlap, is what should be measured.

中文翻译：

---

对忠诚有信心：寻找模型机制时超越电路重叠

最近的许多语言模型 (LM) 可解释性研究都采用了电路框架，该框架旨在找到最小的计算子图或电路，以解释给定任务上的 LM 行为。大多数研究通过对每个边缘独立执行因果干预来确定哪些边缘属于 LM 电路，但这与模型大小的扩展性很差。边缘归因修补（EAP）是一种基于梯度的近似干预措施，已成为解决该问题的可扩展但不完善的解决方案。在本文中，我们介绍了一种新方法 - 集成梯度 EAP（EAP-IG） - 旨在更好地保持电路的核心属性：忠实性。如果可以消除电路外部的所有模型边缘而不改变模型在任务中的性能，则电路是忠实的；忠实性是研究电路而不是完整模型的理由。我们的实验表明，使用 EAP 发现的电路不如使用 EAP-IG 发现的电路忠实，尽管两者都与之前使用因果干预发现的电路有很高的节点重叠。我们更普遍地得出结论，当使用电路来比较用于解决任务的机制模型时，应该衡量的是忠实性，而不是重叠。