[论文] A Framework for Context Enrichment and Selective Robustness in LLM Soc...

论文概要

研究领域: NLP 作者: Mattia Greiche, Reihaneh Rabbany, Maximilian Puelma Touzel 发布时间: 2026-11-04 arXiv: 2611.06914

中文摘要

研究者将大语言模型用作社会模拟中的智能体，通过观察信息和背景信息来引导其行为。然而已有研究表明，LLM 智能体在意义保持的提示变化下会显著改变行为，这给社会模拟带来了验证难题。一个合理假设是：这种脆弱性源于缺乏根基（grounding），可通过丰富上下文缓解。但更丰富的上下文是否真正提升鲁棒性仍不清楚——部分因为上下文丰富度和鲁棒性很少以支持受控比较的方式被定义。我们通过受控上下文丰富化解决第一个需求：一组系统性地丰富不同类型上下文的提示替换规则（中性填充、通用社会背景、描述性决策根基、规范性决策根基）。通过引入选择性鲁棒性解决第二个需求：对意义保持的提示变化不敏感，同时对决策相关的社会世界变化产生方向性响应。我们在类囚徒困境博弈中演示了该框架，展示如何识别哪些上下文丰富化形式能产生更优且更鲁棒的智能体行为。不同形式的丰富化产生了不同行为模式，且在两个测试模型间存在差异。

原文摘要

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 for social simulation. One plausible hypothesis is that this brittleness reflects a lack of grounding and can be reduced by enriching the context provided to the agents. Whether richer context improves robustness remains unclear, in part because context richness and robustness are rarely defined in ways that support controlled comparisons. We address the first need with controlled context enrichment, a set of prompt substitution rules that systematically enriches distinct type...

#论文 #arXiv #NLP #小凯