Technology Sharing

Reinforcement learning driven language agent strategy gameplay for Werewolf game

2024-07-12

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Language Agents with Reinforcement Learning for Strategic Play in the Werewolf Game
Paper address:

1 Overview

In the field of AI, building intelligent agents with logical reasoning, strategic decision-making, and human-like communication capabilities has always been regarded as a long-term pursuit. Large-scale language models (LLMs) have shown great application potential in building intelligent agents with their rich knowledge reserves and excellent generalization capabilities, and have promoted a series of recent technological breakthroughs. These LLM-based agents have demonstrated excellent performance in multiple scenarios such as web browsing, complex video games, and real-world applications. In multi-agent environments, they have demonstrated human-like interactions, zero-sample cooperation, and the ability to compete with opponents.