A Necessary Step toward Faithfulness: Measuring and Improving Consistency in Free-Text Explanations

2025-05-25

AI Summary

Large language models (LLMs) frequently generate free-text explanations for high-stakes AI decisions that lack prediction-explanation consistency (PEX consistency), with over 62% of samples exhibiting inconsistency. Method: We propose the first learnable and evaluable PEX consistency metric, pioneering the integration of evidential weighting into textual explanation consistency modeling. Our approach jointly leverages evidential-weighted consistency quantification, direct preference optimization (DPO), and cross-model-family alignment training across LLaMA, Phi, and Qwen architectures. Contribution/Results: Experiments demonstrate a 43.1%–292.3% improvement in PEX consistency and up to a 9.7% gain in explanation faithfulness. This work establishes the first theoretically grounded and engineering-practical PEX consistency benchmark for explainable AI.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Humans and AI: Explainable AI (XAI) for Human Understanding
• Natural Language Processing: (Large) Language Models

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Faithful free-text explanations are important to ensure transparency in high-stakes AI decision-making contexts, but they are challenging to generate by language models and assess by humans. In this paper, we present a measure for Prediction-EXplanation (PEX) consistency, by extending the concept of weight of evidence. This measure quantifies how much a free-text explanation supports or opposes a prediction, serving as an important aspect of explanation faithfulness. Our analysis reveals that more than 62% explanations generated by large language models lack this consistency. We show that applying direct preference optimization improves the consistency of generated explanations across three model families, with improvement ranging from 43.1% to 292.3%. Furthermore, we demonstrate that optimizing this consistency measure can improve explanation faithfulness by up to 9.7%.

Problem

• Measuring consistency in free-text AI explanations
• Improving explanation faithfulness via direct optimization
• Addressing low consistency in large language model outputs

Innovation

• Measure PEX consistency via weight of evidence
• Apply direct preference optimization for improvement
• Optimize consistency to enhance explanation faithfulness