Instruction tuning has empowered large language models (LLMs) to achieve remarkable performance, yet its success heavily depends on the availability of large-scale, high-quality instruction-response pairs. To meet this demand, various methods have been developed to synthesize data at scale. However, current methods for scaling up data generation often overlook a crucial aspect: the alignment between instructions and responses. We hypothesize that the quality of instruction-response pairs is determined not by the individual quality of each component, but by the degree of mutual alignment. To address this, we propose a Mutual Alignment Framework (MAIN) which enforces coherence between instructions and responses through mutual constraints. We demonstrate that MAIN generalizes well across model architectures and sizes, achieving state-of-the-art performance on LLaMA, Mistral, and Qwen models across diverse benchmarks. This work underscores the critical role of instruction-response alignment in enabling generalizable and high-quality instruction tuning for LLMs. All code is available from our repository.
@article{arxiv.2504.12913,
title = {MAIN: Mutual Alignment Is Necessary for instruction tuning},
author = {Fanyi Yang and Jianfeng Liu and Xin Zhang and Haoyu Liu and Xixin Cao and Yuefeng Zhan and Hao Sun and Weiwei Deng and Feng Sun and Qi Zhang},
journal= {arXiv preprint arXiv:2504.12913},
year = {2025}
}