信息物理系统中LLM代理规划策略的战略评估
原标题：Strategic Evaluation of Planning Strategies for LLM Agents in Cyber-Physical Systems
AI 摘要
该研究提出了一种针对信息物理系统中LLM代理规划策略的受控物理基准测试方法，通过智能电网需求响应系统评估了预定义、顺序、层次和搜索四种执行架构。研究发现，强制搜索在所有基线种子中表现最优，而执行保真度需要超越模式一致性，客观替代虽保持一致性为1.0，但电压短缺增加了2.68倍。在144个场景、576个回合的测试中，应用截止日期可行性可将遗憾值从90.7降至29.0，比固定顺序策略提升61.1%，表明实时可行性应概率性处理。
正文节选
Computer Science > Multiagent Systems Title:Strategic Evaluation of Planning Strategies for LLM Agents in Cyber-Physical Systems View PDF HTML (experimental) Abstract:Evaluations of LLM planning agents largely ask whether a task succeeds or a declared plan is followed. In strategic cyber-physical systems, a stronger question is whether the planning architecture remains appropriate after autonomous participants respond and physics constrains the outcome. We introduce a controlled, phy