@inproceedings{liu-etal-2025-mixture,
    title = "Mixture of insigh{T}ful Experts ({M}o{TE}): The Synergy of Reasoning Chains and Expert Mixtures in Self-Alignment",
    author = "Liu, Zhili  and
      Gou, Yunhao  and
      Chen, Kai  and
      Hong, Lanqing  and
      Gao, Jiahui  and
      Mi, Fei  and
      Zhang, Yu  and
      Li, Zhenguo  and
      Jiang, Xin  and
      Liu, Qun  and
      Kwok, James",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingestion-acl-25/2025.acl-long.151/",
    pages = "3022--3038",
    ISBN = "979-8-89176-251-0",
    abstract = "As the capabilities of large language models (LLMs) continue to expand, aligning these models with human values remains a significant challenge. Recent studies show that reasoning abilities contribute significantly to model safety, while integrating Mixture-of-Experts (MoE) architectures can further enhance alignment.In this work, we address a fundamental question:\textit{How to effectively incorporate reasoning abilitiesand MoE architectures into self-alignment processin LLMs?}We propose Mixture of insighTful Experts (MoTE), a novel framework that synergistically combines reasoning chains and expert mixtures to improve self-alignments.From a data perspective, MoTE employs a structured reasoning chain comprising four key stages: Question Analysis, Answer Guidance, Safe Answer, and Safety Checking. This approach enhances safety through multi-step reasoning and proves effective even for smaller and less powerful LLMs (e.g., 7B models). From an architectural perspective, MoTE adopts a multi-LoRA framework with step-level routing, where each expert is dedicated to a specific reasoning step. This design eliminates the need for balance losses, ensures stable training, and supports adaptive inference lengths. Experimental results demonstrate that MoTE significantly improves model safety, jailbreak resistance, and over-refusal capabilities, achieving performance comparable to OpenAI{'}s state-of-the-art o1 model."
}