Structurally distinct circuits for literal sequence copying across token frequency bands implement the same computation, shown by broad transfer of band-specific edges, a shared core recovering 99% performance, and interchangeable representations via causal interventions.
Mechanistic Interpretability as Statistical Estimation: A Variance Analysis
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
abstract
Mechanistic Interpretability (MI) aims to reverse-engineer model behaviors by identifying functional sub-networks. Yet, the scientific validity of these findings depends on their stability. In this work, we argue that circuit discovery is not a standalone task but a statistical estimation problem built upon causal mediation analysis (CMA). We uncover a fundamental instability at this base layer: exact, single-input CMA scores exhibit high intrinsic variance, implying that the causal effect of a component is a volatile random variable rather than a fixed property. We then demonstrate that circuit discovery pipelines inherit this variance and further amplify it. Fast approximation methods, such as Edge Attribution Patching and its successors, introduce additional estimation noise, while aggregating these noisy scores over datasets leads to fragile structural estimates. Consequently, small perturbations in input data or hyperparameters yield vastly different circuits. We systematically decompose these sources of variance and advocate for more rigorous MI practices, prioritizing statistical robustness and routine reporting of stability metrics.
years
2026 4representative citing papers
CIF certifies interventional interpretability metrics as bounded causal means with anytime-valid confidence sequences, including under adaptive sampling, cutting certification cost 10–30× with betting sequences.
Authors create a benchmark across discrete/continuous and static/dynamical systems and introduce the Causal Abstraction Error (CAE) metric that reliably distinguishes valid from invalid causal abstractions when it includes faithfulness testing.
citing papers explorer
-
Many Circuits, One Mechanism: Input Variation and Evaluation Granularity in Circuit Discovery
Structurally distinct circuits for literal sequence copying across token frequency bands implement the same computation, shown by broad transfer of band-specific edges, a shared core recovering 99% performance, and interchangeable representations via causal interventions.
-
Certified Interventional Fidelity: Anytime-Valid, Adaptive Evaluation of Causal Claims in Mechanistic Interpretability
CIF certifies interventional interpretability metrics as bounded causal means with anytime-valid confidence sequences, including under adaptive sampling, cutting certification cost 10–30× with betting sequences.
-
Validating Causal Abstraction Metrics on Simulated Complex Systems
Authors create a benchmark across discrete/continuous and static/dynamical systems and introduce the Causal Abstraction Error (CAE) metric that reliably distinguishes valid from invalid causal abstractions when it includes faithfulness testing.
- Necessary, Decodable and Reversible, Yet Not Transferable: A Stress Test for Attention-Head Role Claims