Codenames Game Data

该数据集包含了在《代号》游戏中进行的实验数据，这些实验涉及ACE代理与其他多种代理的互动，追踪了在多局游戏中的性能指标。数据不仅包括了ACE代理与静态基础代理合作时的性能统计数据，还涉及了使用不同内部语言模型的对比实验。这些实验比较了ACE代理在与使用相同语言模型的代理合作与不使用时的表现差异。该数据集涵盖了大约85,000局游戏，涉及多组代理的互动。研究任务旨在评估自适应代理在《代号》游戏中的性能。

This dataset includes experimental data collected from gameplay experiments in *Codenames*, which involve interactions between the ACE Agent and various other agents, with performance metrics tracked across numerous game rounds. The data encompasses not only performance statistics for the ACE Agent when collaborating with static baseline agents, but also controlled comparative experiments utilizing distinct internal large language models (LLMs). These experiments compare the performance discrepancies of the ACE Agent when collaborating with agents that employ the same LLM versus when collaborating with agents that do not. This dataset covers approximately 85,000 game rounds involving interactions between multiple groups of agents. The associated research task aims to evaluate the performance of adaptive agents in the game *Codenames*.