多方模拟框架中协作行为的情境敏感性评估
原标题：Evaluating Context Sensitivity of Collaboration Behavior in Multi Party Simulation Frameworks
AI 摘要
该研究针对基于大语言模型（LLM）的多方模拟框架，开展了一项受控的全因子实验，考察团队构成、团队规模（3/5/7/9人）和目标类型（头脑风暴、选择、谈判、执行）等情境因素如何影响协作行为。研究在Sotopia、TinyTroupe和Scripted三个框架下覆盖96种设置，并采用Bales互动过程分析和Meier协作质量框架进行量化。结果显示，模拟器系统性地偏好顺畅协作，而非分歧、谈判压力和协调崩溃，这是首次系统证明情境因素会影响LLM多方模拟中涌现的协作行为。
正文节选
Evaluating Context Sensitivity of Collaboration Behavior in Multi Party Simulation Frameworks Abstract Group Social Behavior is increasingly used as a lens to study Large language model (LLM) simulations to aid model training and evaluation. However, little prior work has examined how contextual factors such as team size and goal type manifest in long running simulations. To overcome this gap, we conduct a controlled, full-factorial study examining how contextual factors affect collaboration be