Anchored Alignment for Self-Explanations Enhancement

TL;DR

This paper proposes an alignment methodology for large language models to improve their self-explanation capabilities without relying on annotated rationales, using novel techniques like Anchor Preference Pairs and tailored preference optimization.

Contribution

It introduces a new alignment approach with Anchor Preference Pairs and tailored strategies, enhancing explanation quality and accuracy in LLMs without annotated data.

Findings

Significant improvement in explanation quality.

Maintains accuracy comparable to other fine-tuning methods.

Effective categorization of model outputs enhances preference selection.

Abstract

In this work, we introduce a methodology for alignment designed to enhance the ability of large language models (LLMs) to articulate their reasoning (self-explanation) even in the absence of annotated rationale explanations. Our alignment methodology comprises three key components: explanation quality assessment, self-instruction dataset generation, and model alignment. Additionally, we present a novel technique called Alignment with Anchor Preference Pairs, which improves the selection of preference pairs by categorizing model outputs into three groups: consistently correct, consistently incorrect, and variable. By applying tailored strategies to each category, we enhance the effectiveness of Direct Preference Optimization (DPO). Our experimental results demonstrate that this approach significantly improves explanation quality while maintaining accuracy compared to other fine-tuning strategies.