Large Language Models are In-context Teachers for Knowledge Reasoning

TL;DR

This paper explores using large language models as in-context teachers for reasoning tasks, demonstrating that self-generated explanations and alignment techniques significantly improve performance over human-crafted examples.

Contribution

It introduces the Self-Explain and Teach-Back methods, showing how LLMs can effectively teach themselves and others, surpassing human teachers in in-context learning accuracy.

Findings

Self-Explain outperforms human-crafted exemplars.

Matching teacher and student explanations improves in-context teaching.

Teach-Back enhances performance, e.g., 7B model surpasses GPT-3.5 in medical QA.

Abstract

In this work, we study in-context teaching (ICT), where a teacher provides in-context example rationales to teach a student to reason over unseen cases. Human teachers are usually required to craft in-context demonstrations, which are costly and have high variance. We ask whether a large language model (LLM) can serve as a more effective in-context teacher for itself or other LLMs, compared to humans. Inspired by the Encoding Specificity Hypothesis from human episodic memory, we hypothesize that in-context exemplars crafted by the teacher should match the training data of the student. This hypothesis motivates us to propose Self-Explain where an LLM's self-elicited explanations are used as in-context demonstrations for prompting it as they are generalized from the model's training examples. Self-Explain is shown to significantly outperform using human-crafted exemplars and other baselines. Furthermore, we reveal that for ICT, rationales from different teacher LLMs or human experts that more resemble the student LLM's self-explanations are better in-context demonstrations. This supports our encoding specificity hypothesis. We then propose Teach-Back that aligns a teacher LLM with the student to enhance the ICT performance. For example, Teach-Back enables a 7B model to teach the much larger GPT-3.5 in context, surpassing human teachers by around 5% in test accuracy on medical question answering.