SciAgent：用于通用科学推理的统一多智能体系统
摘要
最新进展表明，大型语言模型已使 AI 系统在特定科学任务上达到专家水平，但这些系统仍局限于狭义且手工设计。我们提出 SciAgent，一个为通用科学推理而设计的统一多智能体系统——即能够跨学科和难度级别灵活调整推理策略的能力。SciAgent 将问题解决组织为层次化过程：协调器智能体解释每个问题的学科领域和复杂度，动态协调由专业工作系统组成的子智能体，这些子智能体在符号演绎、概念建模、数值计算和验证之间进行交互。这些智能体协作构建并细化针对每个任务的推理管道。在数学和物理奥林匹克竞赛（IMO、IMC、IPhO、CPhO）中，SciAgent 始终达到或超越人类金牌获得者水平，展示了领域的广泛性和推理的可适应性。此外，SciAgent 还在国际化学奥林匹克竞赛（IChO）和人类最后考试（HLE）基准测试中选取的若干问题上进行了测试，进一步证明了该系统在多样化科学领域中的泛化能力。本工作将 SciAgent 确立为通用科学智能——即能够以专家水平进行连贯、跨学科推理的 AI 系统——的具体一步。
引用
@article{arxiv.2511.08151,
title = {SciAgent: A Unified Multi-Agent System for Generalistic Scientific Reasoning},
author = {Xuchen Li and Ruitao Wu and Xuanbo Liu and Xukai Wang and Jinbo Hu and Zhixin Bai and Bohan Zeng and Hao Liang and Leheng Chen and Mingrui Chen and Haitian Zhong and Xuanlin Yang and Xu-Yao Zhang and Liu Liu and Jia Li and Kaiqi Huang and Jiahao Xu and Haitao Mi and Wentao Zhang and Bin Dong},
journal= {arXiv preprint arXiv:2511.08151},
year = {2025}
}
备注
1. To ensure result rigor, the model outputs require further evaluation by human experts. 2. The results may affect our conclusions and methods, thus necessitating a more detailed review. 3. We anticipate subsequent revisions may be substantial, potentially involving major adjustments to the methodology. Given the uncertainty surrounding the revision process, we decide to request a withdrawal