Saved in:
Bibliographic Details
Main Authors: Palumbo, Nils, Mangal, Ravi, Wang, Zifan, Vijayakumar, Saranya, Pasareanu, Corina S., Jha, Somesh
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2407.13594
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909654881140736
author Palumbo, Nils
Mangal, Ravi
Wang, Zifan
Vijayakumar, Saranya
Pasareanu, Corina S.
Jha, Somesh
author_facet Palumbo, Nils
Mangal, Ravi
Wang, Zifan
Vijayakumar, Saranya
Pasareanu, Corina S.
Jha, Somesh
contents Mechanistic interpretability aims to reverse engineer the computation performed by a neural network in terms of its internal components. Although there is a growing body of research on mechanistic interpretation of neural networks, the notion of a mechanistic interpretation itself is often ad-hoc. Inspired by the notion of abstract interpretation from the program analysis literature that aims to develop approximate semantics for programs, we give a set of axioms that formally characterize a mechanistic interpretation as a description that approximately captures the semantics of the neural network under analysis in a compositional manner. We demonstrate the applicability of these axioms for validating mechanistic interpretations on an existing, well-known interpretability study as well as on a new case study involving a Transformer-based model trained to solve the well-known 2-SAT problem.
format Preprint
id arxiv_https___arxiv_org_abs_2407_13594
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Validating Mechanistic Interpretations: An Axiomatic Approach
Palumbo, Nils
Mangal, Ravi
Wang, Zifan
Vijayakumar, Saranya
Pasareanu, Corina S.
Jha, Somesh
Machine Learning
Mechanistic interpretability aims to reverse engineer the computation performed by a neural network in terms of its internal components. Although there is a growing body of research on mechanistic interpretation of neural networks, the notion of a mechanistic interpretation itself is often ad-hoc. Inspired by the notion of abstract interpretation from the program analysis literature that aims to develop approximate semantics for programs, we give a set of axioms that formally characterize a mechanistic interpretation as a description that approximately captures the semantics of the neural network under analysis in a compositional manner. We demonstrate the applicability of these axioms for validating mechanistic interpretations on an existing, well-known interpretability study as well as on a new case study involving a Transformer-based model trained to solve the well-known 2-SAT problem.
title Validating Mechanistic Interpretations: An Axiomatic Approach
topic Machine Learning
url https://arxiv.org/abs/2407.13594