Towards Best Practices of Activation Patching in Language Models: Metrics and Methods
This work addresses methodological inconsistencies for researchers in mechanistic interpretability, offering incremental improvements through best practice recommendations.
The paper tackles the lack of consensus in activation patching for mechanistic interpretability by systematically examining how hyperparameters like evaluation metrics and corruption methods affect results, finding that variations lead to disparate outcomes in localization and circuit discovery tasks.
Mechanistic interpretability seeks to understand the internal mechanisms of machine learning models, where localization -- identifying the important model components -- is a key step. Activation patching, also known as causal tracing or interchange intervention, is a standard technique for this task (Vig et al., 2020), but the literature contains many variants with little consensus on the choice of hyperparameters or methodology. In this work, we systematically examine the impact of methodological details in activation patching, including evaluation metrics and corruption methods. In several settings of localization and circuit discovery in language models, we find that varying these hyperparameters could lead to disparate interpretability results. Backed by empirical observations, we give conceptual arguments for why certain metrics or methods may be preferred. Finally, we provide recommendations for the best practices of activation patching going forwards.