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

Min Zhang (Harbin Institute of Technology, Shenzhen) · Jie Liu (City University of Hong Kong) · Kehai Chen (Harbin Institute of Technology (Shenzhen)) · Liang Yue (Harbin Institute of Technology, Shenzhen) · Yihong Tang (Harbin Institute of Technology)
boost-qacognitive levelscomplex tasksdata augmentationfine-tuning datasethigh-quality data generationinstruction fine-tuningmulti-agent conversationsmultitask generalizationopenhermes2.5orca-math-200kpedagogically grounded teaching scenariospretrained modelsprocqareasoning abilitiesteacher-student interaction

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.