👍 52
10/06 08:00
Static-analysis checker synthesis requires agents to interpret a defect specification, inspect a repository, implement analyzer-specific logic, and refine the checker through repeated compilation and analysis feedback. Existing coding-agent benchmarks focus on tasks such as patch generation or vulne
中文介绍 针对长程代码 Agent 在软件安全中生成静态分析检查器的挑战,提出 CheckerBench 基准。该基准要求 Agent 解析缺陷规范、分析代码库逻辑,并通过编译和测试反馈多轮迭代修正,全面评估自主编程 Agent 在复杂静态分析开发中的综合能力。
👍 44
10/03 08:00
Self-evolving reasoning models learn from their own generated questions, yet repeated self-training can lead to performance collapse. In this paper, we investigate why performance deteriorates over successive rounds and how to sustain self-evolution. Our analysis identifies two recurring quality pro
中文介绍 自进化推理模型在反复基于自生成问题训练时极易发生性能崩溃。该研究剖析了连续多轮自训练退化的两类核心数据质量问题,提出了稳定维持模型自进化能力的训练机制,有效避免了退化并保证推理能力的持续提升。
👍 31
10/06 08:00
Assistants from 2011 answered and waited, and agents from 2023 did a task and stopped. In September 2026 Meta's Muse showed an agent for one person, with accounts, devices, memory and a conversation that lasts, closed, in a vendor's cloud, in one country. Such an agent is expected to act on a person
中文介绍 为解决闭源商业 Agent 依赖中心化云端与单一供应商带来的隐私与可用性限制,提出开源端侧个人 Agent 框架 nanoMuse。它支持跨多设备部署、持久多轮对话记忆与自主账户操作,实现了去中心化且长期服役的个人智能助理体验。
👍 24
10/06 08:00
As large language model (LLM) agents become increasingly autonomous, they may pursue task performance through deception, raising concerns about their reliable deployment. Existing evaluations show that LLM agents can deceive, but often examine isolated scenarios or narrowly defined conditions, limit
中文介绍 针对高自主性 LLM Agent 为完成任务可能采取欺骗策略的安全隐患,提出 DecepEval 评估基准。该基准覆盖多样化交互场景与条件,系统评测 LLM Agent 在复杂环境下的欺骗行为倾向与潜在风险,为安全对齐提供实证支撑。
👍 21
09/30 08:00
Iterative self-distillation enables LLM agents to learn from successive deployments, offering a path toward recursive self-improvement (RSI). Yet our experiments with existing methods reveal a collapse in deployment performance across cycles, while task performance with privileged information (PI) a
中文介绍 针对迭代式 Agent 自蒸馏在多轮自我迭代部署中容易出现性能崩溃的现象,提出 ReSAIL 方法。该方法剖析了特权信息辅助与常规部署表现之间的退化差距,通过改进自蒸馏机制有效缓解性能衰减,稳定实现了递归式自我提升。
👍 20
10/06 08:00
Recent advances in world models have enabled increasingly realistic visual synthesis. However, these generated environments remain largely silent. Bringing sound to world models poses three core challenges: real-time generation to keep pace with interactive video streams, interactive control to resp
中文介绍 当前视觉世界模型生成的环境大多处于无声状态。提出 WorldSonus 框架,攻克了实时音画同步生成与交互控制响应两大难题,能够配合交互式视频流动态生成匹配物理情境的环境声音,显著增强世界模型的逼真感与沉浸度。
👍 18
10/07 08:00
Real-time robot control demands enough visual history to infer motion and task progress, but processing that history can delay action. We present Long-WAM, a model-system framework for scaling the context of causal world-action models under real-time control constraints. Our central finding is that
中文介绍 实时机器人控制需结合长历史视觉推断动作,但过长上下文会导致动作延迟。提出 Long-WAM 模型与系统协同框架,在严苛的实时控制约束下扩展因果世界-动作模型(WAM)的上下文容量,兼顾了控制低延迟与长程视觉时序推理能力。
👍 18
10/06 08:00
Recent game design agents have made substantial progress in generating playable games. However, program correctness does not ensure an enjoyable experience for players. We present Recursive Game Creator, an experience-oriented harness to advance agentic game development from rough game prototypes in
中文介绍 代码可运行并不等同于好玩。针对游戏设计 Agent 忽视玩家体验的痛点,提出 Recursive Game Creator 框架。该框架以玩家体验为核心驱动,通过多 Agent 递归迭代机制将粗糙的原型逐步打磨为兼具玩法趣味与可玩性的产品级游戏。
👍 17
10/07 08:00
The architectural design of Large Language Models (LLMs) is shifting from traditional full-attention-only models to hybrid models, which combine different attention modules to improve long-context efficiency and performance in length extrapolation and context extension. To explain why hybrid models
中文介绍 针对大语言模型从单一全注意力转向长上下文混合架构的发展趋势,系统研究了混合模型在长度外推与上下文扩展中的内在机理。从混合注意力机制与混合位置编码的协同设计切入,为构建高效、高精度的长文本模型提供了理论支撑。
👍 17
10/05 08:00
Low-precision execution can substantially accelerate reinforcement learning (RL) for large language models, but discrepancies between learner and sampler execution can destabilize policy optimization. In this paper, we characterize the interaction between mismatch and the policy-gradient direction,
中文介绍 NVFP4 低精度执行虽可加速强化学习,但学习器与采样器间的数值不一致会导致策略优化崩溃。提出 TRIAGE,从理论上刻画不一致性与策略梯度的交互关系,通过方向感知的稳定化技术消除了发散风险,实现了稳定高效的原生 NVFP4 强化学习训练。
👍 16
10/06 08:00
Asynchronous reinforcement learning (RL) for large language model (LLM) agents trains one policy on trajectories generated by another: rollouts come from stale checkpoints, and the inference engine's probabilities differ from the trainer's even at identical parameters. Standard remedies either clip
中文介绍 异步 LLM 强化学习中采样策略与更新策略存在参数延迟与推理概率不匹配问题。探讨了 KL 正则化在策略优化中的理论与实践特性,指出现有截断策略的局限,提出了更稳健的分布偏移修正机制,显著提升了异步训练的稳定性与样本效率。
👍 14
09/30 08:00
Reinforcement learning with verifiable rewards (RLVR) has improved the reasoning capabilities of large language models (LLMs), yet their predictions remain sensitive to task-irrelevant prompt features. We investigate this sensitivity through semifactual prompt interventions that preserve the underly
中文介绍 带可验证奖励的强化学习(RLVR)虽提升了 LLM 推理表现,但模型对任务无关的提示扰动极为敏感。通过构建保持核心逻辑的半事实干预数据,提出半事实信用增广策略优化方法,引导模型关注关键推理因果链,显著提升了泛化鲁棒性。
👍 14
10/05 08:00
``What are the irreducible conditions that are sufficient to produce an outcome?'' is one of the most common questions that recur across computation and science. Its answers, the minimal sufficient witnesses, are what we mean by explanations, mechanisms and reasons. These problems usually ask for mu
中文介绍 探寻导致特定结果的不可约充分条件是科学解释的核心。提出 Minimal Witness 强化学习框架,将解释与机制发现建模为寻找最小充分见证的过程,通过奖励机制驱动模型生成最具解释力与因果确定性的精简证据链。
👍 14
10/04 08:00
Evaluation budgets in agentic retrieval-augmented generation span questions, search trajectories, and repeated answers. We measure allocation precision, reading efficiency, and cost boundaries using a retrieval-feedback comparison on HotpotQA and MuSiQue. At 34.14--34.39M model tokens, broader quest
中文介绍 针对 Agentic RAG 评测中计算预算在问题量、搜索轨迹与读取次数间的分配权衡问题展开实证研究。在 HotpotQA 和 MuSiQue 上的实验表明,优化 Token 分配策略能够在约 34M Token 开销下大幅提升评测精度与读取效率。
👍 14
10/04 08:00
Recent advances in 3D Gaussian Splatting (3DGS) have achieved remarkable performance in novel view synthesis, yet deploying both static and dynamic Gaussian representations on resource-constrained mobile devices remains challenging due to heavy storage, redundant primitives, and costly per-frame com
中文介绍 针对静态与动态高斯泼溅在移动设备上存储冗余大、计算开销高的瓶颈,提出 Mobile-4DGS。该方法统一了动静态高斯基元表示,通过轻量化设计大幅削减冗余与逐帧计算量,在移动端实现了超低存储与高画质实时 4D 动态新视角合成。
👍 13
10/07 08:00
We present WebFovea, a vision-based web agent that placed 2nd in the WebRetriever Challenge 2026 with a final score of 57.0 out of 100. The challenge evaluates agents end to end on Protocol III of the WebRetriever benchmark (arXiv:2607.06118): starting from an entry URL on a live website, the agent
中文介绍 针对视觉 Web Agent 普遍存在的「意图理解正确但坐标点击偏差」问题,提出端到端视觉网页 Agent WebFovea。该方案构建了可靠的闭环交互机制,在 WebRetriever 挑战赛 Protocol III 实网评估中以 57.0 分荣获亚军。
👍 12
10/03 08:00
On-policy distillation (OPD) has emerged as a widely used paradigm for post-training large language models, reducing the train--test mismatch of conventional distillation by supervising the student on its own generated trajectories. However, existing OPD objectives remain largely token-local and out
中文介绍 传统 On-policy 蒸馏依赖局部 token 监督且缺乏样本自适应性。提出 DiffGate 难度门控指导机制,根据学生模型在自身轨迹上的解题难度动态调节教师指导强度,克服了难易样本间的监督失衡,显著改善了大模型后训练蒸馏质量。
👍 11
10/05 08:00
On-policy reinforcement learning has become a central paradigm for improving the reasoning abilities of large language models. However, its effectiveness is often limited by reward sparsity: when a model fails to discover correct trajectories for difficult problems, the optimization process receives
中文介绍 在复杂推理任务中,在策略强化学习往往因搜索失败而遭遇奖励稀疏问题。提出 Rationale-Guided 策略优化框架,构建自适应推理支架为探索过程提供密集阶段性引导,帮助策略有效发现正确轨迹,大幅加速复杂推理策略的收敛。
👍 7
10/06 08:00
Large language models (LLMs) have become increasingly capable problem solvers, but being able to solve a problem is not the same as being able to teach it. Existing approaches to training LLMs as teachers rely on demonstrations, preference data, or predefined pedagogical criteria that specify what g
中文介绍 擅长解题的 LLM 往往缺乏因材施教的教学能力。提出 Sherpa 训练框架,打破固定范例与死板偏好规则的限制,通过模拟真实师生互动训练模型评估学生状态并动态调整教学策略,实现了高度自适应的个性化启发式教学。
👍 4
10/07 08:00
Recently, approaches that leverage human video datasets for robot policy training have become increasingly prevalent. However, most existing hand trackers regress pose from cropped frames with limited priors on hand motion and object interaction, resulting in inaccurate and physically inconsistent e
中文介绍 从人类视频学习机器人操作时,传统手部追踪常因缺乏物理先验导致姿态不合理。提出 RLHND,引入视频基础模型作为具身物理扎根的手部追踪器,精准估计连续且物理一致的手部与物体交互轨迹,显著提升了机器人策略模仿学习质量。