From Static to Dynamic: Benchmarking Real-World Code Review with MCR-Bench

MCR-Bench基准推动代码审查从静态评估转向动态真实场景。模拟实际开发中的多轮审查交互，覆盖缺陷检测、建议生成等关键任务，更全面地评估模型审查能力，为代码审查智能化研究提供新标准。

📅 08-27📎 arXiv AI Papers

WikiSkill: Compiling Agent Experience into Persistent Knowledge for Skill Evolution

WikiSkill框架将智能体执行任务的经验编译为持久化知识库，实现技能持续进化。通过结构化存储与检索机制使智能体在新任务中复用历史经验，显著提升学习效率与任务表现，为构建可积累知识的自主智能体提供新范式。

📅 08-27📎 arXiv AI Papers

SWE-Prime: Fewer Trajectories, Better Performance

SWE-Prime方法以更少的训练轨迹实现更优性能。通过精选高质量示范轨迹与高效策略优化，在减少数据需求的同时提升代码生成与修复能力，为数据受限场景下的软件工程模型训练提供实用方案。

📅 08-27📎 arXiv AI Papers

RedEvoAgent: Automatic Red-Teaming Agent with Experience-Driven Skill Evolution

RedEvoAgent提出经验驱动的技能进化机制，使红队智能体自动生成对抗攻击以探测AI系统漏洞。通过积累攻击经验动态更新技能库，显著提升攻击覆盖率和效率，为AI安全评估提供自动化解决方案。

📅 08-27📎 arXiv AI Papers

Mechanistic Reaction Prediction via Discrete Flow Matching on Graph-Structured Electron Occupation

提出基于图结构电子占据的离散流匹配方法用于机理级反应预测。将反应机理建模为电子转移过程，利用离散流匹配生成产物图，兼顾化学有效性与预测精度，为合成路线规划提供有力工具。

📅 08-27📎 arXiv AI Papers

Persona-Execution Separation: An Architecture Pattern for Evolving LLM Agents under Execution Audit

Persona-Execution分离架构将LLM智能体的人格设定与执行逻辑解耦，并在执行审计下支持技能持续进化。该模式增强系统可解释性、可控性与安全性，为可审计的自主智能体提供新设计范式。

📅 08-27📎 arXiv AI Papers

Beyond F1: Evaluating Coverage and Failure Recovery in AI Model Security Scanners

Static scanners are increasingly used to identify executable or otherwise unsafe content in machine- learning artifacts, yet conventional evaluation metrics characterize only cases where a scanner yields a usable security judgment. We evaluate ModelScan, ModelAudit, and Fickling using a controlled,

📅 08-27📎 arXiv AI Papers

Learning a Continuous Sepsis Severity Score Without Hour-by-Hour Supervision: A Two-Site Retrospective Study

Currently used sepsis severity indices rely on fixed variables and weights established decades ago, which are coarsely discretized and calibrated to a cohort that no longer reflects contemporary critical care. No alternative learned directly from patient trajectories is in routine use. We conducted

📅 08-27📎 arXiv AI Papers

CLAP: Cross-Embodiment Video World Models are Zero-Shot Physical Simulators

State-of-the-art action-conditioned video models are typically restricted to a single robot embodiment, preventing them from leveraging the vast corpus of heterogeneous video data that contains rich signals for learning generalizable physics. To bridge this gap, we introduce CLAP, a framework for cr

📅 08-27📎 arXiv AI Papers

How Language Models Organize and Structure Moral Knowledge

本文探究语言模型如何组织与结构道德知识，通过分析内部表示揭示模型对道德概念的编码方式。研究发现模型将道德判断映射到可解释的语义空间，并依赖情境与角色信息动态调整道德推理。该工作为理解大模型的价值对齐机制提供新视角，有助于构建更可控、符合伦理的AI系统。

📅 08-27📎 arXiv AI Papers

Planetary prediction engine: Automating global models via Earth AI

谷歌AI推出地球智能自动化全球模型，打造行星预测引擎。该引擎整合多源地球观测数据，通过AI自动构建和优化全球气候、环境等预测模型，大幅提升模拟精度与效率，助力应对气候变化、灾害预警等全球性挑战，为科学决策提供智能支撑。

📅 08-27📎 Google AI Blog

Teamwork: When AI Becomes a Research Partner

探讨AI作为研究伙伴的协作模式，通过案例展示AI深度参与文献分析、数据挖掘与假设验证。文章强调人机协同能显著提升研究效率与创新质量，分析分工边界与协作伦理，为科研团队提供人机协作的可行路径与新视角。

📎 Antigravity Blog