SCOTT: Self-Consistent Chain-of-Thought Distillation

TL;DR

This paper introduces SCOTT, a method for distilling large language models into smaller, self-consistent models that generate more faithful rationales for their predictions, improving interpretability without sacrificing performance.

Contribution

The paper presents a novel knowledge distillation approach that ensures the smaller model produces consistent and faithful chain-of-thought rationales aligned with the teacher model.

Findings

Smaller models can generate more faithful rationales using SCOTT.

SCOTT achieves comparable task performance to larger models.

The method enhances the interpretability and trustworthiness of LLMs.

Abstract

Large language models (LMs) beyond a certain scale, demonstrate the emergent capability of generating free-text rationales for their predictions via chain-of-thought (CoT) prompting. While CoT can yield dramatically improved performance, such gains are only observed for sufficiently large LMs. Even more concerning, there is little guarantee that the generated rationales are consistent with LM's predictions or faithfully justify the decisions. In this work, we propose a faithful knowledge distillation method to learn a small, self-consistent CoT model from a teacher model that is orders of magnitude larger. To form better supervision, we elicit rationales supporting the gold answers from a large LM (teacher) by contrastive decoding, which encourages the teacher to generate tokens that become more plausible only when the answer is considered. To ensure faithful distillation, we use the teacher-generated rationales to learn a student LM with a counterfactual reasoning objective, which prevents the student from ignoring the rationales to make inconsistent predictions. Experiments show that, while yielding comparable end-task performance, our method can generate CoT rationales that are more faithful than baselines do. Further analysis suggests that such a model respects the rationales more when making decisions; thus, we can improve its performance more by refining its rationales.