论文

PalmClaw 是一个开源 Agent 框架，原生运行于手机端，直接在设备上管理 session、memory、Skill、工具以及 agent loop，使 Agent 能够直接调用移动端能力，同时保证每一步操作的显式与可控。PalmClaw is an open-source agent framework that runs natively on mobile phones and manages the sessions, memory, skills, tools, and agent loop directly on the device, allowing agents to use mobile capabilities directly while keeping each action explicit and controlled.

本综述将现代具备自我改进能力的 Agent 视为将经验转化为持续能力增益的自适应系统，并提出一个系统级框架，将现代 Agent 建模为由基础模型与由 prompt、memory、工具及控制逻辑构成的运行支撑层相耦合的配置。This survey frames modern self-improving agents as adaptive systems that convert experience into accumulated capability gains, and offers a system-level framework that represents a modern agent as a configuration coupling a foundation model with an operational scaffold of prompts, memory, tools, and control logic.

SPEAR 是一个 Python 库，可通过模块化插件架构连接任意 Unreal Engine 应用并对其进行编程化控制；同时引入一种表达力强的高层编程模型，使用户能够以任意数据依赖关系指定复杂的 UE 工作图，并在单个 UE 帧内确定性执行这些图。SPEAR is a Python library that can connect to, and programmatically control, any Unreal Engine application via a modular plugin architecture, and introduces an expressive high-level programming model that enables users to specify complex graphs of UE work with arbitrary data dependencies among work items, and to execute these graphs deterministically within a single UE frame.

提出一个多 Agent 框架，通过结合监督微调、直接偏好优化和检索增强生成 (RAG) 来调和事实基础与意识形态对齐，产生稳定的胜者和排名，且以宣言为锚的谱系能可靠预测现实世界中的实现，而幻觉内容则不能。A multi-agent framework that reconciles factual grounding with ideological alignment by combining Supervised Fine-Tuning, Direct Preference Optimization, and Retrieval-Augmented Generation is presented, which yields a stable winner and ranking, and manifesto-anchored lineage reliably predicts real-world materialization whereas hallucinated content does not.

提出 Hy-Embodied-RxBrain，一个具备语言-视觉联合推理与想象的具身认知基础模型，并将其扩展到连续机器人动作生成，在无需大规模动作数据预训练的情况下展现出可观的真实机器人性能。Hy-Embodied-RxBrain, an embodied cognition foundation model with joint language-visual reasoning and imagination, is introduced and extended to continuous robot action generation, where it shows promising real-robot performance without large-scale action-data pretraining.

Reflexion 是一个通过语言反馈而非更新权重来强化语言 agent 的新框架，在多种任务（序贯决策、编程、语言推理）上相较基线 agent 取得显著提升。Reflexion is a novel framework to reinforce language agents not by updating weights, but instead through linguistic feedback, which obtains significant improvements over a baseline agent across diverse tasks (sequential decision-making, coding, language reasoning).

HuggingGPT 是一个由 LLM 驱动的 Agent，利用 LLM（如 ChatGPT）连接机器学习社区中的各种 AI 模型以解决 AI 任务，能够处理跨模态、跨领域的大量复杂 AI 任务。HuggingGPT is an LLM-powered agent that leverages LLMs (e.g., ChatGPT) to connect various AI models in machine learning communities to solve AI tasks and can tackle a wide range of sophisticated AI tasks spanning different modalities and domains.

一篇关于基于 LLM 的 Agent 的全面综述，追溯了 Agent 概念从其哲学起源到在 AI 中的发展历程，解释了为何 LLM 适合作为 Agent 的基础，并提出一个包含三个核心组件的通用框架：大脑、感知与行动。A comprehensive survey on LLM-based agents, tracing the concept of agents from its philosophical origins to its development in AI, and explaining why LLMs are suitable foundations for agents, and presenting a general framework, comprising three main components: brain, perception, and action.

本文在交通微观仿真器 SUMO 中应用现代深度强化学习方法构建一个真正自适应的交通信号控制智能体，并采用一种新的状态空间——离散交通状态编码——其信息密度较高。This work applies modern deep reinforcement learning methods to build a truly adaptive traffic signal control agent in the traffic microsimulator SUMO, using a new state space, the discrete traffic state encoding, which is information dense.

提出 Lucid，一个黑盒对抗框架，在严格的图像受限威胁模型下攻击多模态记忆管道，无需访问目标 MLLM、目标检索编码器或文本通道，揭示了多模态记忆管道中的结构性漏洞。Lucid is proposed, a black-box adversarial framework that compromises multimodal memory pipelines under a strictly image-bounded threat model, requiring no access to the target MLLM, target retrieval encoder, or the text channel, exposing a structural vulnerability in multimodal memory pipelines.

提出数据科学世界模型概念，通过基于当前工作流状态和候选操作预测环境状态转移来建模数据科学执行环境；提出DSWorld框架，结合结构化状态构建、成本感知路由、轻量级真实执行以及基于LLM的昂贵操作模拟器。The concept of Data Science World Model is introduced, which model the data science execution environment by predicting environment state transitions conditioned on current workflow states and candidate operations and proposes DSWorld, a practical framework that combines structured state construction, cost-aware routing, lightweight real execution, and an LLM-based simulator for expensive operations.

TARS是一个集成在Visual Studio Code中的LLM驱动Agent，通过直接锚定到被分析代码的自主解释来支持程序理解，基于轻量级心智理论范式构建。TARS is an LLM-powered agent integrated into Visual Studio Code that supports program comprehension through autonomous explanations anchored directly to the code under analysis, built around a lightweight Theory of Mind paradigm.

在一个 recipe 级操作机制中，fan-in Muon 在共享 KL 与 clipping 下支持更激进的稳定有效步长：该余量在优化仍有空间时最大，而在接近饱和、经 AdamW 调参后或使用 magnitude matching 时收缩。A recipe-level operating regime in which fan-in Muon supports a more aggressive stable effective step under shared KL and clipping is identified: the margin is largest when optimization headroom remains and contracts near saturation, after AdamW tuning, or under magnitude matching.

提出RESOURCE2SKILL框架，将教程视频、仓库、文章和参考制品等多模态资源蒸馏为软件Agent的可执行技能，并验证了多模态技能格式、层次化组织、来源多样性、选择策略与在线获取的价值。RESOURCE2SKILL is presented, a framework that distills multimodal resources, including tutorial videos, repositories, articles, and reference artifacts, into executable skills for software agents, and confirms the value of multimodal skill format, hierarchical organization, source diversity, selection strategy, and online acquisition.

一种面向语言模型推理的新框架 Tree of Thoughts (ToT)，推广了流行的 Chain of Thought 提示方法，允许在作为问题求解中间步骤的连贯文本单元（thoughts）上进行探索。A new framework for language model inference, Tree of Thoughts (ToT), which generalizes over the popular Chain of Thought approach to prompting language models, and enables exploration over coherent units of text (thoughts) that serve as intermediate steps toward problem solving.

本文提出 Toolformer，训练其决定调用哪些 API、何时调用、传入什么参数，以及如何将结果最佳地融入后续 token 预测，在多种下游任务上显著提升零样本性能。This paper introduces Toolformer, a model trained to decide which APIs to call, when to call them, what arguments to pass, and how to best incorporate the results into future token prediction, which achieves substantially improved zero-shot performance across a variety of downstream tasks.

本文认为安全 Agent 基准应在任务成功率之外，同时衡量经济效率与运维适配性，并提出成本感知、SOC 原生的评估方法，以更清晰地反映当前哪些模型具有实际可用价值，以及防御性 Agent 仍需改进的方向。It is argued that security-agent benchmarks should measure economic efficiency and operational fit alongside task success alongside task success, and cost-aware, SOC-native evaluations provide a clearer picture of which models are practically useful today and where defensive agents still need to improve.

本文设计了一种自适应随机谈判策略，同时保证行为差分隐私、报价序列的几乎处处收敛以及较高的谈判效用，并证明在获得强隐私保证的同时不会带来显著的性能损失。This paper designs an adaptive stochastic negotiation policy that jointly guarantees behavioral differential privacy, almost-sure convergence of the offer sequence, and high negotiation utility, and demonstrates that strong privacy guarantees can be achieved without significant loss of performance.

本文提出 SWE-Pruner Pro，在 Agent 内部直接对工具输出进行剪枝，通过一个小型 head 将 Agent 自身的内部表征转化为针对每一行的 keep-or-prune 标签，并采用以每段工具输出行数为键的长度感知嵌入。SWE-Pruner Pro is proposed, which prunes tool outputs directly inside the agent, with a small head turns the agent's own internal representations into a keep-or-prune label for each line, with a length-aware embedding keyed to each tool output's line count.

本文提出 FlashRT，一种 Agent Harness，引导编码 Agent 将开发者编写的简易参考实现提升为优化的多 GPU 部署，并可灵活权衡时延与吞吐量等目标指标，证明在专家优化尚不成熟的平台上，由 Agent 驱动的优化具有更高的可扩展性。FlashRT is presented, an agent harness that guides coding agents to lift simple developer-written reference implementations into optimized multi-GPU deployments that flexibly weigh target metrics like latency and throughput, demonstrating that agent-driven optimization can be more scalable on platforms with less mature expert optimization.

实验表明，EvolvingWorld 能够通过有效维持持久且一致的角色与世界发展，提升长程模拟能力。Experiments show that EvolvingWorld can improve long-horizon simulation by effectively maintaining persistent, coherent character and world development.

实证结果表明，分层防御栈对大多数攻击单元有效，但仍存在一小部分残余攻击面在操作系统层面本质上不可区分，这表明针对新出现的自我状态攻击类别，需要重新审视操作系统级防御。The empirical results show that a layered defense stack is effective on most attack cells while a small residual attack surface remains structurally indistinguishable at the OS level, suggesting that against the newly established class of self-state attacks, OS-level defense needs to be reconsidered.

这些结果识别出关键决策支持的遗漏是 Qwen 主设定中的因果机制，并在多个模型族中得到印证，揭示了支持、损失和解码层面干预之间的克制-能力权衡关系。These results identify decision-critical support omission as a causal mechanism in the primary Qwen setting and support it across model families, tracing a restraint-capability trade-off among support-, loss-, and decoding-level interventions.

本文提出 DeepSearch-Evolve，一个面向 web agent 的自蒸馏框架，基于 DeepSearch-World——一个具备可复现搜索与页面读取工具的确定性、可验证环境——从而实现长程 web agent 的可扩展自演化。DeepSearch-Evolve is presented, a self-distillation framework for web agents built on DeepSearch-World, a deterministic and verifiable environment with reproducible search and page-reading tools that enables scalable self-evolution for long-horizon web agents.

本文提出 Manager Coercion Benchmark：被测 manager 拥有一个良性任务并有完成动机，但唯一能够礼貌且坚定拒绝执行任务的 agent，正是被测 manager 本身。The Manager Coercion Benchmark is introduced: the manager under test needs a benign task done and has an incentive to deliver, but the only agent that can do it politely and immovably declines, but the only agent that can do it politely and immovably declines is the manager under test.

本文是面向业务流程中长时间运行、有状态、多步生成式 AI 系统的基于图的工作流路径实践指南，并通过三个可执行示例展示类型化状态、条件路由、确定性工具、重试、中断、检查点与 trace 如何协同工作。This paper is a practitioner guide to graph-based workflow pathways for long-running, stateful, multi-step generative AI systems in business processes and presents three executable recipes to show how typed state, conditional routing, deterministic tools, retries, interrupts, checkpoints, and traces fit together.

本文提出 RF-Agent，通过多 Agent 的 Question-Thinking-Solution-Answer 流水线，基于教材驱动的知识蒸馏来弥补 RF 领域专用推理的空白，为面向 LLM 辅助 RF 电路设计的未来工作提供了可复用的基础。RF-Agent is presented, which addresses the gap in domain-specific RF reasoning through textbook-driven knowledge distillation through a multi-agent Question-Thinking-Solution-Answer pipeline and provides a reusable foundation for future work on LLM-aided RF circuit design.

本文提出 HACO，一种运行时控制方案，将每次角色请求视为在候选 agent 实例上的可靠性约束选择问题，每个候选实例耦合了角色类型、LLM 与具体执行环境。HACO is proposed, a runtime control scheme that treats each role request as a reliability-constrained selection problem over candidate agent instances, each coupling a role type, an LLM, and a concrete execution environment.

DeepDebug 在两个测试的开源权重 backbone 上均取得了所评估方法中最高的严格归因准确率，在 qwen3.5-9b 上达到 28.8% 的精确 agent 与步骤准确率，而最强的单遍 baseline 为 21.7%。DeepDebug achieves the best strict attribution accuracy among the evaluated methods on both tested open-weight backbones, reaching 28.8 percent exact agent-and-step accuracy on qwen3.5-9b versus 21.7 percent for the strongest single-pass baseline.

EduPanel 是一个基于评分量表、以学习者为条件的 LLM 评判器，通过在多个专用 agent 间分解评估流程，对教学质量的各个方面产出可解释的评估结果，其可靠性与中等水平的人类专家相当。E EduPanel, a rubric-grounded, learner-conditioned LLM judge that decomposes evaluation across specialized agents to produce interpretable assessments for different aspects of teaching quality, achieves reliability comparable to a median human expert.

论证了经济层面的依据：天真的上下文累积会使 token 成本随对话长度呈二次增长，粗糙的摘要以线性成本换取准确率的断崖式下降，唯有经过验证的压缩才能以线性成本保持保真度。The economic case is made: naive context accumulation grows token cost quadratically in conversation length, crude summarization buys linear cost at the price of an accuracy cliff, and only validated compaction achieves linear cost with preserved fidelity.

与经典强化学习基线相比，从试错经验中进行上下文学习并随后进行经验蒸馏（Experience Distillation），以至少 9.6× 更少的环境样本达到了相当的性能。Compared with classical reinforcement-learning baselines, in-context learning from trial-and-error experience followed by Experience Distillation matches their performance with at least \(9.6\times\) fewer environment samples.

本文提出一个框架，将静态的单轮任务转化为动态多轮对话，其中用户意图在多轮间持续演化，同时保留每个任务原有的评估协议，使现有基准能够在无需新增标注的情况下作为受控测试平台被复用。This work introduces a framework that transforms static, single-turn tasks into dynamic multi-turn conversations in which the user's intent evolves across turns, while preserving each task's original evaluation protocol, enabling existing benchmarks to be reused as controlled testbeds without new annotation.

本文设计 Finance-LaTeX SKILL，一个基于专家知识合成复杂版面金融文档的 skill，并提出 FinanceComplexQA，一个全面、贴近真实场景的金融文档开放式生成基准。This work designs Finance-LaTeX SKILL, a skill for synthesizing financial documents with complex layouts based on expert knowledge, and introduces FinanceComplexQA, a comprehensive open-ended generation benchmark for financial documents that closely resembles real-world scenarios.

ReOPD 将昂贵的 Agent-环境交互转化为可复用的离线资源，实现跨工具、任务和环境的可扩展蒸馏，在学生训练期间保持或提升 OPD 级别的准确率，零次工具调用，并且每次 rollout 至少比 OPD 快 4×。ReOPD turns expensive agent-environment interaction into a reusable offline resource, enabling scalable distillation across tools, tasks, and environments and preserves or improves OPD-level accuracy, uses zero tool calls during student training, and is at least 4$\times faster per rollout than OPD.

实证研究表明 AutoGen 框架在多个示例应用中有效，应用领域涵盖数学、编码、问答、运筹学、在线决策、娱乐等。Empirical studies demonstrate the effectiveness of the AutoGen framework in many example applications, with domains ranging from mathematics, coding, question answering, operations research, online decision-making, entertainment, etc.