MASTER: Enhancing Large Language Model via Multi-Agent Simulated Teaching

TL;DR

MASTER introduces a multi-agent simulation approach to generate high-quality instruction data, significantly improving large language models' reasoning and multitask performance without extensive data collection.

Contribution

The paper presents a novel multi-agent interaction method for data augmentation, enhancing instruction fine-tuning datasets and model capabilities in NLP tasks.

Findings

Models fine-tuned with BOOST-QA outperform baselines on multiple benchmarks.

MASTER significantly boosts reasoning abilities in complex tasks.

The approach reduces data collection costs while improving model generalization.

Abstract

Instruction fine-tuning is crucial in NLP tasks, enhancing pretrained models' instruction-following capabilities and task-specific performance. However, obtaining high-quality fine-tuning data for large models is challenging due to data collection difficulties and high production costs. To address this, we propose MASTER, a novel data augmentation method that enriches original data through interactions among multiple agents with varying cognitive levels. We simulate three pedagogically grounded teaching scenarios, leveraging multi-agent conversations to generate high-quality teacher-student interaction data. Utilizing MASTER, we construct BOOST-QA, a fine-tuning dataset augmented from existing datasets like Orca-Math-200k, ProcQA, and OpenHermes2.5. Experiments show that models fine-tuned with BOOST-QA perform excellently across multiple benchmarks, demonstrating strong multitask generalization. Notably, MASTER significantly improves models' reasoning abilities in complex tasks, providing valuable insights for future research.