LLM社会智能体的上下文丰富与选择性鲁棒性框架
原标题：A Framework for Context Enrichment and Selective Robustness in LLM Social Agents
AI 摘要
研究者提出一个用于LLM社会智能体的上下文丰富与选择性鲁棒性框架，将上下文丰富分解为中性填充、通用社会背景、描述性决策依据和规范性决策依据四类条件，并引入“选择性鲁棒性”作为验证标准，要求智能体对保持语义的提示变化不敏感、但对社会世界中与决策相关的实质变化有方向性响应。作者在类似囚徒困境的博弈中做了初步实验，发现不同丰富方式产生不同行为模式，且两个被测模型表现各异。该工作针对LLM社会模拟中的验证难题，代码与数据已开源。
正文节选
SocialAgent: Second Workshop on Large Language Models for Social Reasoning and Simulation A Framework for Context Enrichment and Selective Robustness in LLM Social Agents Abstract Researchers use large language models as agents in social simulations, where agents are prompted with observations and background information intended to shape their actions. Yet LLM agents have been shown to substantially change their behavior under meaning-preserving prompt variations, creating a validation problem f