Enhancing Consistency of Werewolf AI through Dialogue Summarization and Persona Information
arXiv:2603.07111v1 Announce Type: new Abstract: The Werewolf Game is a communication game where players' reasoning and discussion skills are essential. In this study, we present a Werewolf AI agent developed for the AIWolfDial 2024 shared task, co-hosted with the 17th INLG. In recent years, large language models like ChatGPT have garnered attention for their exceptional response generation and reasoning capabilities. We thus develop the LLM-based agents for the Werewolf Game. This study aims to enhance the consistency of the agent's utterances by utilizing dialogue summaries generated by LLMs and manually designed personas and utterance examples. By analyzing self-match game logs, we demonstrate that the agent's utterances are contextually consistent and that the character, including tone, is maintained throughout the game.
arXiv:2603.07111v1 Announce Type: new Abstract: The Werewolf Game is a communication game where players' reasoning and discussion skills are essential. In this study, we present a Werewolf AI agent developed for the AIWolfDial 2024 shared task, co-hosted with the 17th INLG. In recent years, large language models like ChatGPT have garnered attention for their exceptional response generation and reasoning capabilities. We thus develop the LLM-based agents for the Werewolf Game. This study aims to enhance the consistency of the agent's utterances by utilizing dialogue summaries generated by LLMs and manually designed personas and utterance examples. By analyzing self-match game logs, we demonstrate that the agent's utterances are contextually consistent and that the character, including tone, is maintained throughout the game.
Executive Summary
This article presents a novel approach to enhancing the consistency of a Werewolf AI agent through dialogue summarization and persona information. The authors develop a large language model (LLM)-based agent for the AIWolfDial 2024 shared task, leveraging the capabilities of ChatGPT and its ilk. By incorporating dialogue summaries generated by LLMs and manually designed personas/utterance examples, the agent's utterances are rendered contextually consistent and maintain a consistent tone throughout the game. The authors demonstrate the effectiveness of their approach through self-match game logs analysis. The study contributes to the development of more sophisticated AI agents for communication games, with potential applications in areas such as education, entertainment, and human-computer interaction.
Key Points
- ▸ The authors develop an LLM-based Werewolf AI agent for the AIWolfDial 2024 shared task.
- ▸ Dialogue summarization and persona information are used to enhance the agent's consistency.
- ▸ The approach demonstrates contextually consistent utterances and maintained tone throughout the game.
Merits
Contribution to AI Agent Development
The study presents a novel approach to enhancing the consistency of AI agents in communication games, contributing to the development of more sophisticated AI agents.
Potential Applications
The approach has potential applications in areas such as education, entertainment, and human-computer interaction, where AI agents can engage in more natural and effective communication with humans.
Demerits
Limited Generalizability
The study focuses on a specific game (Werewolf) and may not be directly applicable to other communication games or domains.
Dependence on LLMs
The approach relies on the performance and limitations of large language models, which may introduce biases and errors in the agent's responses.
Expert Commentary
The study presents a promising approach to enhancing the consistency of AI agents in communication games. However, the approach relies heavily on the performance and limitations of large language models, which may introduce biases and errors in the agent's responses. Additionally, the study's focus on a specific game (Werewolf) may limit the generalizability of the findings. Nonetheless, the study contributes to the development of more sophisticated AI agents and has potential applications in areas such as education, entertainment, and human-computer interaction.
Recommendations
- ✓ Future studies should investigate the application of the approach to other communication games or domains to improve generalizability.
- ✓ The use of more diverse and representative datasets is recommended to mitigate potential biases and errors introduced by large language models.