Abstract: As large language models (LLMs) continue to demonstrate impressive reasoning capabilities, LLM- based multi-agent has become an increasingly compelling area of research. Despite the potential, the field faces a notable gap: the scarcity of LLM-based simulators tailored for realistic, multi-agent interactions. Most existing multi-agent simulators are missing textual interfaces and quantitative evaluation metrics, limiting the assessment of complex interactions between agents. Our proposed simulator,$\texttt{AutoDiner}$, replicates a detailed restaurant management scenario requiring advanced communication and teamwork among agents, providing a uniquely realistic and complex research environment. $\texttt{AutoDiner}$ not only fosters intricate agent interactions but also incorporates varying levels of difficulty and performance metrics for comprehensive benchmarking. These features make $\texttt{AutoDiner}$ an exemplary platform for advancing the understanding and capabilities of LLM-based agents in navigating complex tasks and enhancing cooperative strategies in realistic settings.
Paper Type: short
Research Area: NLP Applications
Contribution Types: Publicly available software and/or pre-trained models
Languages Studied: English
0 Replies
Loading