FitCF: A Framework for Automatic Feature Importance-guided Counterfactual Example Generation

TL;DR

This paper introduces FitCF, a novel framework that uses feature attribution methods to automatically generate high-quality counterfactual examples for NLP, enhancing model interpretability and data augmentation.

Contribution

The paper proposes FitCF, a new framework that verifies and uses feature attribution to generate effective counterfactuals, outperforming existing methods in quality and reliability.

Findings

FitCF outperforms two state-of-the-art baselines.

LIME and Integrated Gradients are effective attribution methods within FitCF.

Number of demonstrations significantly impacts performance.

Abstract

Counterfactual examples are widely used in natural language processing (NLP) as valuable data to improve models, and in explainable artificial intelligence (XAI) to understand model behavior. The automated generation of counterfactual examples remains a challenging task even for large language models (LLMs), despite their impressive performance on many tasks. In this paper, we first introduce ZeroCF, a faithful approach for leveraging important words derived from feature attribution methods to generate counterfactual examples in a zero-shot setting. Second, we present a new framework, FitCF, which further verifies aforementioned counterfactuals by label flip verification and then inserts them as demonstrations for few-shot prompting, outperforming two state-of-the-art baselines. Through ablation studies, we identify the importance of each of FitCF's core components in improving the quality of counterfactuals, as assessed through flip rate, perplexity, and similarity measures. Furthermore, we show the effectiveness of LIME and Integrated Gradients as backbone attribution methods for FitCF and find that the number of demonstrations has the largest effect on performance. Finally, we reveal a strong correlation between the faithfulness of feature attribution scores and the quality of generated counterfactuals, which we hope will serve as an important finding for future research in this direction.