Learning from Mistakes via Cooperative Study Assistant for Large Language Models

TL;DR

This paper introduces SALAM, a cooperative framework with an auxiliary agent that helps large language models learn from their mistakes, significantly improving their accuracy by leveraging interactive guidance and mistake analysis.

Contribution

The paper proposes SALAM, a novel cooperative framework with an auxiliary agent that enhances LLM learning from mistakes through interaction and imitation learning, improving accuracy.

Findings

SALAM improves LLM accuracy by up to 6.6 on BBH.

SALAM improves LLM accuracy by up to 12.6 on BBQ.

The framework effectively assists LLMs in error analysis and correction.

Abstract

Large language models (LLMs) have demonstrated their potential to refine their generation based on their own feedback. However, the feedback from LLM itself is often inaccurate, thereby limiting its benefits. In this paper, we propose Study Assistant for Large LAnguage Model (SALAM), a novel framework with an auxiliary agent to assist the main LLM in learning from mistakes through interactive cooperation. In the gathering phase, the student assistant agent probes the main LLM, analyzes its errors, and collects the interaction in a mistake memory. During the examination phase, the study assistant provides guidelines by retrieving relevant cases to help the main LLM anticipate and avoid similar errors. We first investigate the effectiveness of a general study assistant and then customize it to provide LLM-specific guidance through imitation learning from successful guidance experiences. Our experiments on three LLMs using two challenging frameworks demonstrate that SALAM can significantly boost LLMs by an accuracy margin of up to 6.6 on BBH and 12.6 on BBQ.