Argument-Based Consistency in Toxicity Explanations of LLMs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mothilal, Ramaravind Kommiya, Roy, Joanna, Ahmed, Syed Ishtiaque, Guha, Shion
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908787446644736
author Mothilal, Ramaravind Kommiya
Roy, Joanna
Ahmed, Syed Ishtiaque
Guha, Shion
author_facet Mothilal, Ramaravind Kommiya
Roy, Joanna
Ahmed, Syed Ishtiaque
Guha, Shion
contents The discourse around toxicity and LLMs in NLP largely revolves around detection tasks. This work shifts the focus to evaluating LLMs' reasoning about toxicity - from their explanations that justify a stance - to enhance their trustworthiness in downstream tasks. Despite extensive research on explainability, it is not straightforward to adopt existing methods to evaluate free-form toxicity explanation due to their over-reliance on input text perturbations, among other challenges. To account for these, we propose a novel, theoretically-grounded multi-dimensional criterion, Argument-based Consistency (ArC), that measures the extent to which LLMs' free-form toxicity explanations reflect an ideal and logical argumentation process. Based on uncertainty quantification, we develop six metrics for ArC to comprehensively evaluate the (in)consistencies in LLMs' toxicity explanations. We conduct several experiments on three Llama models (of size up to 70B) and an 8B Ministral model on five diverse toxicity datasets. Our results show that while LLMs generate plausible explanations to simple prompts, their reasoning about toxicity breaks down when prompted about the nuanced relations between the complete set of reasons, the individual reasons, and their toxicity stances, resulting in inconsistent and irrelevant responses. We open-source our code (https://github.com/uofthcdslab/ArC) and LLM-generated explanations (https://huggingface.co/collections/uofthcdslab/arc) for future works.
format Preprint
id arxiv_https___arxiv_org_abs_2506_19113
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Argument-Based Consistency in Toxicity Explanations of LLMs
Mothilal, Ramaravind Kommiya
Roy, Joanna
Ahmed, Syed Ishtiaque
Guha, Shion
Computation and Language
The discourse around toxicity and LLMs in NLP largely revolves around detection tasks. This work shifts the focus to evaluating LLMs' reasoning about toxicity - from their explanations that justify a stance - to enhance their trustworthiness in downstream tasks. Despite extensive research on explainability, it is not straightforward to adopt existing methods to evaluate free-form toxicity explanation due to their over-reliance on input text perturbations, among other challenges. To account for these, we propose a novel, theoretically-grounded multi-dimensional criterion, Argument-based Consistency (ArC), that measures the extent to which LLMs' free-form toxicity explanations reflect an ideal and logical argumentation process. Based on uncertainty quantification, we develop six metrics for ArC to comprehensively evaluate the (in)consistencies in LLMs' toxicity explanations. We conduct several experiments on three Llama models (of size up to 70B) and an 8B Ministral model on five diverse toxicity datasets. Our results show that while LLMs generate plausible explanations to simple prompts, their reasoning about toxicity breaks down when prompted about the nuanced relations between the complete set of reasons, the individual reasons, and their toxicity stances, resulting in inconsistent and irrelevant responses. We open-source our code (https://github.com/uofthcdslab/ArC) and LLM-generated explanations (https://huggingface.co/collections/uofthcdslab/arc) for future works.
title Argument-Based Consistency in Toxicity Explanations of LLMs
topic Computation and Language
url https://arxiv.org/abs/2506.19113