@inproceedings{gondo-etal-2024-verification,
    title = "Verification of Reasoning Ability using {BDI} Logic and Large Language Model in {AIW}olf",
    author = "Gondo, Hiraku  and
      Sakaji, Hiroki  and
      Noda, Itsuki",
    editor = "Kano, Yoshinobu",
    booktitle = "Proceedings of the 2nd International AIWolfDial Workshop",
    month = sep,
    year = "2024",
    address = "Tokyo, Japan",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/add-emnlp-2024-awards/2024.aiwolfdial-1.5/",
    pages = "40--47",
    abstract = "We attempt to improve the reasoning capability of LLMs in werewolf game by combining BDI logic with LLMs. While LLMs such as ChatGPT has been developed and used for various tasks, there remain several weakness of the LLMs. Logical reasoning is one of such weakness. Therefore, we try to introduce BDI logic-based prompts to verify the logical reasoning ability of LLMs in dialogue of werewofl game. Experiments and evaluations were conducted using {\textquotedblleft}AI-Werewolf,{\textquotedblright} a communication game for AI with incomplete information. From the results of the game played by five agents, we compare the logical reasoning ability of LLMs by using the win rate and the vote rate against werewolf."
}