Enhance Reasoning for Large Language Models in the Game Werewolf

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wu, Shuang, Zhu, Liwen, Yang, Tao, Xu, Shiwei, Fu, Qiang, Wei, Yang, Fu, Haobo
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914733946306560
author Wu, Shuang
Zhu, Liwen
Yang, Tao
Xu, Shiwei
Fu, Qiang
Wei, Yang
Fu, Haobo
author_facet Wu, Shuang
Zhu, Liwen
Yang, Tao
Xu, Shiwei
Fu, Qiang
Wei, Yang
Fu, Haobo
contents This paper presents an innovative framework that integrates Large Language Models (LLMs) with an external Thinker module to enhance the reasoning capabilities of LLM-based agents. Unlike augmenting LLMs with prompt engineering, Thinker directly harnesses knowledge from databases and employs various optimization techniques. The framework forms a reasoning hierarchy where LLMs handle intuitive System-1 tasks such as natural language processing, while the Thinker focuses on cognitive System-2 tasks that require complex logical analysis and domain-specific knowledge. Our framework is presented using a 9-player Werewolf game that demands dual-system reasoning. We introduce a communication protocol between LLMs and the Thinker, and train the Thinker using data from 18800 human sessions and reinforcement learning. Experiments demonstrate the framework's effectiveness in deductive reasoning, speech generation, and online game evaluation. Additionally, we fine-tune a 6B LLM to surpass GPT4 when integrated with the Thinker. This paper also contributes the largest dataset for social deduction games to date.
format Preprint
id arxiv_https___arxiv_org_abs_2402_02330
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Enhance Reasoning for Large Language Models in the Game Werewolf
Wu, Shuang
Zhu, Liwen
Yang, Tao
Xu, Shiwei
Fu, Qiang
Wei, Yang
Fu, Haobo
Artificial Intelligence
Computation and Language
This paper presents an innovative framework that integrates Large Language Models (LLMs) with an external Thinker module to enhance the reasoning capabilities of LLM-based agents. Unlike augmenting LLMs with prompt engineering, Thinker directly harnesses knowledge from databases and employs various optimization techniques. The framework forms a reasoning hierarchy where LLMs handle intuitive System-1 tasks such as natural language processing, while the Thinker focuses on cognitive System-2 tasks that require complex logical analysis and domain-specific knowledge. Our framework is presented using a 9-player Werewolf game that demands dual-system reasoning. We introduce a communication protocol between LLMs and the Thinker, and train the Thinker using data from 18800 human sessions and reinforcement learning. Experiments demonstrate the framework's effectiveness in deductive reasoning, speech generation, and online game evaluation. Additionally, we fine-tune a 6B LLM to surpass GPT4 when integrated with the Thinker. This paper also contributes the largest dataset for social deduction games to date.
title Enhance Reasoning for Large Language Models in the Game Werewolf
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2402.02330