Local Explanations and Self-Explanations for Assessing Faithfulness in black-box LLMs

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Fragkathoulas, Christos, Chlapanis, Odysseas S.
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910615763681280
author Fragkathoulas, Christos
Chlapanis, Odysseas S.
author_facet Fragkathoulas, Christos
Chlapanis, Odysseas S.
contents This paper introduces a novel task to assess the faithfulness of large language models (LLMs) using local perturbations and self-explanations. Many LLMs often require additional context to answer certain questions correctly. For this purpose, we propose a new efficient alternative explainability technique, inspired by the commonly used leave-one-out approach. Using this approach, we identify the sufficient and necessary parts for the LLM to generate correct answers, serving as explanations. We propose a metric for assessing faithfulness that compares these crucial parts with the self-explanations of the model. Using the Natural Questions dataset, we validate our approach, demonstrating its effectiveness in explaining model decisions and assessing faithfulness.
format Preprint
id arxiv_https___arxiv_org_abs_2409_13764
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Local Explanations and Self-Explanations for Assessing Faithfulness in black-box LLMs
Fragkathoulas, Christos
Chlapanis, Odysseas S.
Computation and Language
Artificial Intelligence
This paper introduces a novel task to assess the faithfulness of large language models (LLMs) using local perturbations and self-explanations. Many LLMs often require additional context to answer certain questions correctly. For this purpose, we propose a new efficient alternative explainability technique, inspired by the commonly used leave-one-out approach. Using this approach, we identify the sufficient and necessary parts for the LLM to generate correct answers, serving as explanations. We propose a metric for assessing faithfulness that compares these crucial parts with the self-explanations of the model. Using the Natural Questions dataset, we validate our approach, demonstrating its effectiveness in explaining model decisions and assessing faithfulness.
title Local Explanations and Self-Explanations for Assessing Faithfulness in black-box LLMs
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2409.13764