Recent LLM agents have made great use of chain of thought reasoning and function calling. As their capabilities grow, an important question arises: can this software represent not only a smart problem-solving tool, but an entity in its own right, that can plan, design immediate tasks, and reason toward broader, more ambiguous goals? To study this question, we adopt an open-ended experimental setting where we augment a pretrained LLM agent with the ability to generate its own tasks, accumulate knowledge, and interact extensively with its environment. We study the resulting open-ended agent qualitatively. It can reliably follow complex multi-step instructions, store and reuse information across runs, and propose and solve its own tasks, though it remains sensitive to prompt design, prone to repetitive task generation, and unable to form self-representations. These findings illustrate both the promise and current limits of adapting pretrained LLMs toward open-endedness, and point to future directions for training agents to manage memory, explore productively, and pursue abstract long-term goals.
ScaffoldAgent: Utility-Guided Dynamic Outline Optimization for Open-Ended Deep Research
National Engineering Research Center of Software Engineering, Peking University(北京大学软件工程国家工程研究中心);School of Computer Science, Peking University(北京大学计算机学院);Key Laboratory of High Confidence Software Technologies, Ministry of Education(教育部高可信软件技术重点实验室);GRG Banking Equipment Co., Ltd.(广电运通金融电子股份有限公司);Center on Frontiers of Computing Studies, Peking University(北京大学计算前沿研究中心);Peking University Information Technology Institute (Tianjin Binhai)(北京大学(天津滨海)信息技术研究院)
A Definition of Open-Ended Learning Problems for Goal-Conditioned Agents
Sorbonne Université(索邦大学);CNRS(法国国家科学研究中心);Institut des Systèmes Intelligents et de Robotique(智能系统与机器人研究所);INRIA Bordeaux Sud-Ouest(法国国家信息与自动化研究所波尔多南部-西部分部);Institute of Cognitive Sciences and Technologies(认知科学与技术研究所);CNR(意大利国家研究委员会);CITIC(信息与通信技术中心);Universidade da Coruña(拉科鲁尼亚大学);Massachusetts Institute of Technology(麻省理工学院)
Distribution Preference Optimization: A Fine-grained Perspective for LLM Unlearning
Tsinghua University(清华大学);The Hong Kong University of Science and Technology(香港科技大学);Jianghuai Advanced Technology Center(江淮先进技术中心);University of Technology Sydney(悉尼大学)
DERA: Enhancing Large Language Model Completions with Dialog-Enabled Resolving Agents
Curai Health(Curai健康)
大语言模型(LLM)已经成为许多自然语言理解任务的有价值的工具。在医疗保健等安全关键型应用中,这些模型的实用性取决于它们生成真实准确和完整的输出的能力。在这项工作中,我们提出了对话使能解析代理(DERA)。DERA是一个范例,通过LLM的会话能力的增加而成为可能,即GPT-4。它为模型提供了一个简单的、可解释的论坛,以交流反馈并迭代地改进输出。我们框架我们的对话作为两个代理类型之间的讨论-一个研究员,谁处理信息,并确定关键的问题组件,和一个决策者,谁拥有自主权,以整合研究员的信息,并作出判断的最终输出。我们测试DERA对三个临床重点任务。对于医疗会话摘要和护理计划生成,DERA在人类专家偏好评估和定量指标方面都显示出比基本GPT-4性能的显着改进。在一项新的发现中,我们还表明,GPT-4在MedQA问答(QA)数据集(Jin et al. 2021,USMLE)的开放式版本上的性能(70%)远高于及格水平(60%),DERA表现出类似的性能。我们在https://github.com/curai/curai-research/tree/main/DERA上发布开放式MEDQA数据集。
A Dual Perspective on Synthetic Trajectory Generators: Utility Framework and Privacy Vulnerabilities
Orange Research(Orange研究院);Université Marie et Louis Pasteur(玛丽-路易斯·巴斯德大学);CNRS(法国国家科学研究中心);institut FEMTO-ST (UMR 6174)(FEMTO-ST研究所);F-25000 Besançon(贝桑松);France(法国)