While most efforts to improve LLM-based tool-using agents focus on the agent itself - through larger models, better prompting, or fine-tuning - agent performance increasingly plateaus due to the quality of the tool interfaces these agents consume. Tool descriptions are often written for human developers and tolerate ambiguity that agents cannot resolve, particularly as the number of candidate tools grows. Existing approaches to improving tool interfaces (1) require re-running a multi-stage per-tool pipeline - synthesizing queries, executing an agent to collect trajectories, annotating trajectories, and prompting a strong LLM multiple times - for every API that enters the catalog, and (2) typically optimize each tool independently, limiting scalability and generalization to unseen tools. We propose Trace-Free+, a curriculum learning framework that progressively transfers supervision from trace-rich settings to trace-free deployment, encouraging the model to internalize reusable patterns of what makes a tool description effective. To support this approach, we construct a large-scale dataset of high-quality tool interfaces derived from real-world APIs through a principled data synthesis workflow. Experiments on widely adopted benchmarks show that Trace-Free+ improves robustness as tool catalogs scale to 150+ candidates - in scaling experiments, reducing accuracy degradation by 29.23% and improving average query-level success by 60.89% on StableToolBench - generalizes across domains without retraining, and provides complementary gains on top of agent fine-tuning.
QUITE: A Query Rewrite System Beyond Rules with LLM Agents
Sichuan University(四川大学);Cornell University(康奈尔大学);Purdue University(普渡大学);Hong Kong University of Science and Technology(香港科学与技术大学);Computer Network Information Center, Chinese Academy of Sciences(中国科学院计算机网络信息中心)
Self-Evolving Agents as Dynamic Graph Transformation: A Survey and New Perspective
School of Computer Science and Engineering, The University of New South Wales(新南威尔士大学计算机科学与工程学院);Faculty of Engineering and Information Technology, University of Technology Sydney(悉尼科技大学工程与信息技术学院);School of Data Science, The Chinese University of Hong Kong-Shenzhen(香港中文大学(深圳)数据科学学院);Zhejiang Gongshang University(浙江工商大学)
AgenticMath: Enhancing LLM Reasoning via Agentic-based Math Data Generation
King’s College London(伦敦国王学院);Hong Kong Baptist University(香港 Baptist 大学);The Hong Kong University of Science and Technology (Guangzhou)(香港科学与技术大学(广州));ByteDance Seed(字节跳动种子);City University of Hong Kong(香港城市大学);Xi’an Jiaotong University(西安交通大学)
GPTSee: Enhancing Moment Retrieval and Highlight Detection via Description-Based Similarity Features
School of Physics and Electronics, Hubei Normal University(湖北师范大学物理与电子科学学院);School of Electronic Science and Engineering, Nanjing University(南京大学电子科学与工程学院)
Matching Matters: A Fair Quality-Efficiency Benchmark for Command-Line Agents
Computer Network Information Center, Chinese Academy of Sciences(中国科学院计算机网络信息中心);Hangzhou Institute for Advanced Study, Chinese Academy of Sciences(中国科学院杭州高等研究院)
Measuring and Reducing Unusable Replies of LLM Agents
East China Normal University(华东师范大学);The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州));Shanghai Artificial Intelligence Laboratory(上海人工智能实验室);Institute of Deep Perception Technology, JITRI(JITRI深度感知技术研究院)
Center for Artificial Intelligence Research (CAiRE)(人工智能研究中心(CAiRE));The Hong Kong University of Science and Technology(香港科技大学)
现有的个性化对话模型使用人类设计的人物描述来提高对话的一致性。从现有的对话中收集这些描述是昂贵的,需要手工制作的功能设计。在本文中,我们提出扩展模型不可知元学习(MAML)(Finn等人,2017年)到个性化的对话学习,而不使用任何人物角色描述。我们的模型通过仅利用从同一用户收集的几个对话样本来学习快速适应新的人物角色,这与根据人物角色描述来调节响应有根本的不同。Persona-chat数据集的经验结果(Zhang et al.,2018)表明,我们的解决方案使用自动评估指标,在人工评估的流畅性和一致性方面优于非元学习基线。
Learning Rhetorical Structure Theory-based descriptions of observed behaviour
Instituto de Telecomunicações (IT-IUL)(电信研究所(IT-IUL));Instituto Universitário de Lisboa (ISCTE-IUL)(里斯本大学学院(ISCTE-IUL));Instituto de Engenharia de Sistemas e Computadores (INESC-ID)(系统与计算机工程研究所(INESC-ID))
在以前的论文中,我们已经提出了一组概念,公理图式和算法,可用于代理学习描述他们的行为,目标,能力和环境。目前的论文提出了一组新的概念,公理模式和算法,允许代理学习新的描述观察到的行为(例如,令人困惑的动作),其参与者(例如,不期望的提议或动作),以及其环境(例如,不相容的命题)。每个学习的描述(例如,某个动作防止将来执行另一个动作)由实体(命题或动作)之间的关系表示,并且由代理仅通过观察、使用域独立公理模式和/或学习算法来学习。代理人用来表示他们所学习的描述的关系受到修辞结构理论(Theory of Rhetorical Structure,简称RBT)的启发。本文的主要贡献是关系家族虽然,启发了对关系让步。族关系的精确定义涉及到一组道义概念,给出了它们的定义和相应的算法。家庭的关系虽然,一旦从代理人的观察中提取,表示惊讶的观察到的行为,并在某些情况下,提出了一个理由。该文件显示的结果,提出的建议,在一个示范方案,使用实现的软件。