Everything, Everywhere, All at Once: Is Mechanistic Interpretability Identifiable?
As AI systems are used in high-stakes applications, ensuring interpretability is crucial. Mechanistic Interpretability (MI) aims to reverse-engineer neural networks by extracting human-understandable algorithms to explain their behavior. This work examines a key question: for a given behavior, and under MI's criteria, does a unique explanation exist? Drawing on identifiability in statistics, where parameters are uniquely inferred under specific assumptions, we explore the identifiability of MI explanations. We identify two main MI strategies: (1) "where-then-what," which isolates a circuit replicating model behavior before interpreting it, and (2) "what-then-where," which starts with candidate algorithms and searches for neural activation subspaces implementing them, using causal alignment. We test both strategies on Boolean functions and small multi-layer perceptrons, fully enumerating candidate explanations. Our experiments reveal systematic non-identifiability: multiple circuits can replicate behavior, a circuit can have multiple interpretations, several algorithms can align with the network, and one algorithm can align with different subspaces. Is uniqueness necessary? A pragmatic approach may require only predictive and manipulability standards. If uniqueness is essential for understanding, stricter criteria may be needed. We also reference the inner interpretability framework, which validates explanations through multiple criteria. This work contributes to defining explanation standards in AI.
Code (1)
Tasks
AllMethods 이 논문이 사용한 방법론
Similar Papers 제목 키워드 기반
LLM Interpretability with Identifiable Temporal-Instantaneous Representation
Despite Large Language Models' remarkable capabilities, understanding their internal representations remains challenging. Mechanistic interpretability tools such as sparse autoencoders (SAEs) were developed to extract in…
Representation LearningIdentifiability and interpretability of hybrid, gray-box models
Model identifiability concerns the uniqueness of uncertain model parameters to be estimated from available process data and is often thought of as a prerequisite for physical interpretability of the parameters in a mecha…
Segment Everything Everywhere All at Once
In this work, we present SEEM, a promptable and interactive model for segmenting everything everywhere all at once in an image, as shown in Fig.1. In SEEM, we propose a novel decoding mechanism that enables diverse promp…
AllDecoderImage SegmentationInteractive Segmentation+5Evaluating the Robustness of the "Ensemble Everything Everywhere" Defense
Ensemble everything everywhere is a defense to adversarial examples that was recently proposed to make image classifiers robust. This defense works by ensembling a model's intermediate representations at multiple noisy i…
Adversarial AttackRobust classificationMechanistic Interpretability for AI Safety -- A Review
Understanding AI systems' inner workings is critical for ensuring value alignment and safety. This review explores mechanistic interpretability: reverse engineering the computational mechanisms and representations learne…