Can LLMs Learn by Teaching for Better Reasoning? A Preliminary Study

TL;DR

This paper explores whether large language models can improve their reasoning by teaching other models, demonstrating preliminary methods that enhance accuracy and generalization without extensive retraining.

Contribution

It introduces three novel methods for LLMs to learn by teaching, enabling reasoning improvements through feedback and iterative learning without additional training.

Findings

Teaching materials with clearer logic improve in-context learning

LbT can help strengthen weak models by teaching stronger ones

Teaching multiple students enhances learning diversity and effectiveness

Abstract

Teaching to improve student models (e.g., knowledge distillation) is an extensively studied methodology in LLMs. However, for humans, teaching improves not only students but also teachers, by fostering more rigorous and clear reasoning as well as knowledge building. We ask: Can LLMs also learn by teaching (LbT) for better reasoning? If the answer is yes, we can potentially unlock the possibility of continuously advancing the models without solely relying on human-produced data or stronger models. In this paper, we provide a preliminary exploration on this question. We show that LbT ideas can be incorporated into existing LLM training/prompting pipelines and bring improvements. Specifically, we design three methods, each mimicking one of the three levels of LbT: observing students' feedback, learning from the feedback, and learning iteratively, with the goals of improving answer accuracy without training or improving models' inherent capability with fine-tuning. We reveal some findings: (1) Teaching materials that make it easier for students to learn have clearer and more accurate logic when using in-context learning as the student's "learning" method; (2) Weak-to-strong generalization: LbT might help improve strong models by teaching weak models; (3) Diversity in students might help: teaching multiple students could be better than teaching one student or the teacher itself. We hope that our exploration can inspire future research on LbT and more broadly adopting the advanced techniques in education to improve LLMs. The code and website are at https://github.com/imagination-research/lbt and https://sites.google.com/view/llm-learning-by-teaching.