本文提出智能体系统的自动化设计(ADAS)新领域,通过 Meta Agent Search 算法让元智能体用代码迭代编程发现新智能体,实验证明其发明的智能体性能优于手工设计且具跨领域鲁棒性。
摘要:
研究人员正在投入大量精力开发强大的通用智能体,其中 Foundation Models 被用作智能体系统内的模块(例如 Chain-of-Thought、Self-Reflection、Toolformer)。然而,机器学习的历史告诉我们,手工设计的解决方案最终会被学习到的解决方案所取代。我们描述了一个新形成的研究领域,即智能体系统的自动化设计(ADAS),旨在自动创建强大的智能体系统设计,包括发明新的构建模块和/或以新的方式组合它们。我们进一步证明,在 ADAS 中存在一种尚未探索但很有前景的方法,即智能体可以用代码定义,并且可以通过元智能体在代码中编写越来越好的智能体来自动发现新智能体。鉴于编程语言是图灵完备的,这种方法在理论上能够学习任何可能的智能体系统:包括新颖的提示、工具使用、工作流程及其组合。我们提出了一种简单而有效的算法 Meta Agent Search 来证明这一想法,其中元智能体基于不断增长的先前发现档案,迭代地编写有趣的新智能体。通过对包括编程、科学和数学在内的多个领域进行广泛实验,我们表明我们的算法可以逐步发明具有新颖设计的智能体,其性能大大优于最先进的手工设计智能体。重要的是,我们始终观察到令人惊讶的结果,即由 Meta Agent Search 发明的智能体即使在不同领域和模型之间迁移时也能保持卓越的性能,证明了它们的鲁棒性和通用性。只要我们安全地开发它,我们的工作就展示了一个令人兴奋的新研究方向的潜力,即自动设计更强大的智能体系统以造福人类。
英文摘要:
Researchers are investing substantial effort in developing powerful general-purpose agents, wherein Foundation Models are used as modules within agentic systems (e.g. Chain-of-Thought, Self-Reflection, Toolformer). However, the history of machine learning teaches us that hand-designed solutions are eventually replaced by learned solutions. We describe a newly forming research area, Automated Design of Agentic Systems (ADAS), which aims to automatically create powerful agentic system designs, including inventing novel building blocks and/or combining them in new ways. We further demonstrate that there is an unexplored yet promising approach within ADAS where agents can be defined in code and new agents can be automatically discovered by a meta agent programming ever better ones in code. Given that programming languages are Turing Complete, this approach theoretically enables the learning of any possible agentic system: including novel prompts, tool use, workflows, and combinations thereof. We present a simple yet effective algorithm named Meta Agent Search to demonstrate this idea, where a meta agent iteratively programs interesting new agents based on an ever-growing archive of previous discoveries. Through extensive experiments across multiple domains including coding, science, and math, we show that our algorithm can progressively invent agents with novel designs that greatly outperform state-of-the-art hand-designed agents. Importantly, we consistently observe the surprising result that agents invented by Meta Agent Search maintain superior performance even when transferred across domains and models, demonstrating their robustness and generality. Provided we develop it safely, our work illustrates the potential of an exciting new research direction toward automatically designing ever-more powerful agentic systems to benefit humanity.
A Sober Look at Agentic Misalignment in Automated Workflows
University of Virginia(弗吉尼亚大学);Georgia Institute of Technology(佐治亚理工学院);University of Utah(犹他大学);University of Notre Dame(圣约翰大学);George Washington University(乔治华盛顿大学)
AutoRAS: Learning Robust Agentic Systems with Primitive Representations
Beijing University of Posts and Telecommunications(北京邮电大学);Guangxi Transportation Science and Technology Group Co., Ltd.(广西交通科技集团有限公司);Fudan University(复旦大学)
HyEvo: Self-Evolving Hybrid Agentic Workflows for Efficient Reasoning
East China Normal University(华东师范大学);Beihang University(北京航空航天大学);Shanghai University of International Business and Economics(上海对外经贸大学);Fudan University(复旦大学)
TopoAgent: An Agentic Framework for Automated Topology Learning in Medical Imaging
Dept. of Computer Science and Engineering, University of Notre Dame(内布拉斯加大学达灵顿分校计算机科学与工程系);Dept. of Computer Science, The University of Texas Rio Grande Valley(德克萨斯大学里奥格兰德谷分校计算机科学系)
Limits to Verification and Validation of Agentic Behavior
代理行为的验证和确认已被建议作为重要的研究重点,以减少与创建通用人工智能相关的风险(Russell et al 2015)。在本文中,我们质疑使用语言的确定性方面的努力,以管理这种风险的适当性。我们首先建立了一个非常一般的形式主义来描述代理行为和可接受的行为的标准。我们表明,确定一个代理是否符合任何特定的标准是不可计算的。我们讨论了与人工证明和自动化行为治理验证相关的负担程度。我们表明,为了确保行为标准本身的可判定性,必须进一步限制代理的能力。然后,我们证明,如果我们的关注涉及到在物理世界的结果,验证的尝试是徒劳的。最后,我们表明,分层架构旨在使这些挑战易于处理的错误等同于意图与行动或结果,从而无法提供任何保证。最后,我们讨论了为什么确定性语言应该从关于一般人工智能安全性的对话中根除。
Composing Verifiable Conceptual Models via Building Blocks: Towards Design-Time Verification of Agentic AI Workflows
Team EVERGREEN Inria Centre Inria d’Université Côte d’Azur(法国国家信息与自动化研究所蔚蓝海岸大学中心EVERGREEN团队);Department of Data Science William & Mary(威廉与玛丽学院数据科学系);Office of Enterprise Research and Innovation Old Dominion University(欧道明大学企业研究与创新办公室)
RefGrader: Automated Grading of Mathematical Competition Proofs using Agentic Workflows
Pennsylvania State University(宾夕法尼亚州立大学);City University of New York(纽约市立大学);New York University(纽约大学);Amirkabir University of Technology(阿米尔卡比尔理工大学);Autodesk(欧特克公司);Carnegie Mellon University(卡内基梅隆大学)