EvalxNLP: A Framework for Benchmarking Post-Hoc Explainability Methods on NLP Models

TL;DR

EvalxNLP is a comprehensive Python framework that benchmarks and evaluates various explainability methods for transformer-based NLP models, aiding stakeholders in selecting suitable explanations.

Contribution

It introduces a unified platform integrating eight explainability techniques and human evaluation, facilitating systematic comparison and advancement of XAI methods in NLP.

Findings

High user satisfaction with EvalxNLP's explanations

Effective benchmarking of multiple explainability methods

Enhanced understanding of explanation properties in NLP

Abstract

As Natural Language Processing (NLP) models continue to evolve and become integral to high-stakes applications, ensuring their interpretability remains a critical challenge. Given the growing variety of explainability methods and diverse stakeholder requirements, frameworks that help stakeholders select appropriate explanations tailored to their specific use cases are increasingly important. To address this need, we introduce EvalxNLP, a Python framework for benchmarking state-of-the-art feature attribution methods for transformer-based NLP models. EvalxNLP integrates eight widely recognized explainability techniques from the Explainable AI (XAI) literature, enabling users to generate and evaluate explanations based on key properties such as faithfulness, plausibility, and complexity. Our framework also provides interactive, LLM-based textual explanations, facilitating user understanding of the generated explanations and evaluation outcomes. Human evaluation results indicate high user satisfaction with EvalxNLP, suggesting it is a promising framework for benchmarking explanation methods across diverse user groups. By offering a user-friendly and extensible platform, EvalxNLP aims at democratizing explainability tools and supporting the systematic comparison and advancement of XAI techniques in NLP.