Saved in:
Bibliographic Details
Main Authors: La Malfa, Emanuele, Vadillo, Jon, Molinari, Marco, Wooldridge, Michael
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.12421
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911209763110912
author La Malfa, Emanuele
Vadillo, Jon
Molinari, Marco
Wooldridge, Michael
author_facet La Malfa, Emanuele
Vadillo, Jon
Molinari, Marco
Wooldridge, Michael
contents This paper introduces a formal notion of fixed point explanations, inspired by the "why regress" principle, to assess, through recursive applications, the stability of the interplay between a model and its explainer. Fixed point explanations satisfy properties like minimality, stability, and faithfulness, revealing hidden model behaviours and explanatory weaknesses. We define convergence conditions for several classes of explainers, from feature-based to mechanistic tools like Sparse AutoEncoders, and we report quantitative and qualitative results for several datasets and models, including LLMs such as Llama-3.3-70B.
format Preprint
id arxiv_https___arxiv_org_abs_2505_12421
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Fixed Point Explainability
La Malfa, Emanuele
Vadillo, Jon
Molinari, Marco
Wooldridge, Michael
Machine Learning
Artificial Intelligence
This paper introduces a formal notion of fixed point explanations, inspired by the "why regress" principle, to assess, through recursive applications, the stability of the interplay between a model and its explainer. Fixed point explanations satisfy properties like minimality, stability, and faithfulness, revealing hidden model behaviours and explanatory weaknesses. We define convergence conditions for several classes of explainers, from feature-based to mechanistic tools like Sparse AutoEncoders, and we report quantitative and qualitative results for several datasets and models, including LLMs such as Llama-3.3-70B.
title Fixed Point Explainability
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2505.12421