BattleAgentBench
收藏资源简介:
BattleAgentBench是由清华大学知识工程组开发的一个用于评估语言模型在多智能体系统中合作与竞争能力的数据集。该数据集设计了三个难度级别和七个阶段,涵盖了从单智能体到多智能体的复杂场景。数据集内容包括游戏规则理解、空间感知、合作与竞争等多种能力评估。创建过程中,研究团队通过设计不同难度的游戏环境和任务,对语言模型进行了细致的评估。该数据集主要应用于人工智能领域,特别是在多智能体系统的合作与竞争能力评估中,旨在提升模型在复杂环境中的适应性和性能。
BattleAgentBench is a dataset developed by the Knowledge Engineering Group of Tsinghua University for evaluating the cooperative and competitive capabilities of language models in multi-agent systems. This dataset is designed with three difficulty levels and seven stages, covering complex scenarios ranging from single-agent to multi-agent settings. It includes evaluations of various capabilities such as game rule comprehension, spatial perception, cooperation and competition. During its development, the research team conducted rigorous evaluations of language models by designing game environments and tasks with varying difficulty levels. This dataset is primarily applied in the field of artificial intelligence, especially for evaluating cooperative and competitive capabilities in multi-agent systems, aiming to enhance the adaptability and performance of models in complex environments.

- 1BattleAgentBench: A Benchmark for Evaluating Cooperation and Competition Capabilities of Language Models in Multi-Agent Systems清华大学知识工程组(KEG) · 2024年



