A Positive Case for Faithfulness: LLM Self-Explanations Help Predict Model Behavior

TL;DR

This paper introduces a new metric, Normalized Simulatability Gain (NSG), to evaluate the faithfulness of LLM self-explanations, demonstrating they significantly improve the prediction of model behavior across various models and datasets.

Contribution

The paper proposes NSG, a scalable metric for assessing explanation faithfulness, and provides empirical evidence that self-explanations enhance model behavior prediction, surpassing external explanations.

Findings

Self-explanations improve behavior prediction by 11-37% NSG.

Self-generated explanations outperform external explanations in predictive power.

5-15% of self-explanations are egregiously misleading.

Abstract

LLM self-explanations are often presented as a promising tool for AI oversight, yet their faithfulness to the model's true reasoning process is poorly understood. Existing faithfulness metrics have critical limitations, typically relying on identifying unfaithfulness via adversarial prompting or detecting reasoning errors. These methods overlook the predictive value of explanations. We introduce Normalized Simulatability Gain (NSG), a general and scalable metric based on the idea that a faithful explanation should allow an observer to learn a model's decision-making criteria, and thus better predict its behavior on related inputs. We evaluate 18 frontier proprietary and open-weight models, e.g., Gemini 3, GPT-5.2, and Claude 4.5, on 7,000 counterfactuals from popular datasets covering health, business, and ethics. We find self-explanations substantially improve prediction of model behavior (11-37% NSG). Self-explanations also provide more predictive information than explanations generated by external models, even when those models are stronger. This implies an advantage from self-knowledge that external explanation methods cannot replicate. Our approach also reveals that, across models, 5-15% of self-explanations are egregiously misleading. Despite their imperfections, we show a positive case for self-explanations: they encode information that helps predict model behavior.