Enregistré dans:
Détails bibliographiques
Auteurs principaux: Chen, Ryan, Ko, Youngmin, Zhang, Zeyu, Cho, Catherine, Chung, Sunny, Giuffré, Mauro, Shung, Dennis L., Stadie, Bradly C.
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2505.11772
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866914506929602560
author Chen, Ryan
Ko, Youngmin
Zhang, Zeyu
Cho, Catherine
Chung, Sunny
Giuffré, Mauro
Shung, Dennis L.
Stadie, Bradly C.
author_facet Chen, Ryan
Ko, Youngmin
Zhang, Zeyu
Cho, Catherine
Chung, Sunny
Giuffré, Mauro
Shung, Dennis L.
Stadie, Bradly C.
contents We introduce LAMP (Local Attribution Mapping Probe), a method that shines light onto a black-box language model's decision surface and studies how reliably a model maps its stated reasons to its reported predictions by approximating a decision surface. LAMP treats the model's own self-reported explanations as a coordinate system and fits a locally linear surrogate that links those weights to the model's output. By doing so, it reveals how much the stated factors steer the model's decisions. We apply LAMP to three tasks: sentiment analysis, controversial-topic detection, and safety-prompt auditing. Across these tasks, LAMP reveals that many language models' locally approximated linear decision landscapes overall agree with human judgments on explanation quality and, on a clinical case-file data set, align with expert assessments. Since LAMP operates without requiring access to model gradients, logits, or internal activations, it serves as a practical and lightweight framework for auditing proprietary language models, and enabling assessment of whether a model appears to behave consistently with the explanations it provides.
format Preprint
id arxiv_https___arxiv_org_abs_2505_11772
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LAMP: Extracting Local Decision Surfaces From Large Language Models
Chen, Ryan
Ko, Youngmin
Zhang, Zeyu
Cho, Catherine
Chung, Sunny
Giuffré, Mauro
Shung, Dennis L.
Stadie, Bradly C.
Machine Learning
We introduce LAMP (Local Attribution Mapping Probe), a method that shines light onto a black-box language model's decision surface and studies how reliably a model maps its stated reasons to its reported predictions by approximating a decision surface. LAMP treats the model's own self-reported explanations as a coordinate system and fits a locally linear surrogate that links those weights to the model's output. By doing so, it reveals how much the stated factors steer the model's decisions. We apply LAMP to three tasks: sentiment analysis, controversial-topic detection, and safety-prompt auditing. Across these tasks, LAMP reveals that many language models' locally approximated linear decision landscapes overall agree with human judgments on explanation quality and, on a clinical case-file data set, align with expert assessments. Since LAMP operates without requiring access to model gradients, logits, or internal activations, it serves as a practical and lightweight framework for auditing proprietary language models, and enabling assessment of whether a model appears to behave consistently with the explanations it provides.
title LAMP: Extracting Local Decision Surfaces From Large Language Models
topic Machine Learning
url https://arxiv.org/abs/2505.11772