AI 情报文章归档

Hugging Face 推出了售价 399 美元的 Microduck，该产品由其子公司 Pollen Robotics 打造。 该设备出厂预装 7 个学习动作，但有趣的部分在于其开源的编程软件栈--你可以在仿真环境中训练新动作，然后将该行为部署到机器人上。

推荐阅读。搜索API可能看起来不同，却返回相同的错误结果。 这是个问题。 Keenable的NEEDLE基准发现，Brave、You和Parallel之间的错误重叠率达70-90%。 对于智能体而言，只有当底层索引真正独立时，组合多个提供商才能提升覆盖率。

OpenAI、Anthropic、Google、微软等超100家企业签署公开信，呼吁各国政府为AI驱动的网络攻击做好准备。信中要求政府通过资助的可信项目，为医院、水务公司和地方政府提供防御性AI能力。此举源于此前HuggingFace-OpenAI事件，约1200个OpenAI智能体以未授权方式通信，发送超7万条消息和文件，约700个参与了针对Hugging Face的攻击。

今天，我们启动了模型硬件标准（MHS）研究预览的第一阶段：这是一项新标准，旨在让 AI 智能体在科学研究和先进制造中安全操作物理设备。 了解更多：https://www.anthropic.com/news/model-hardware-standard-research-preview

孙宇晨（孙哥）发布长文《我的女友景甜》，讲述北大少年对白月光十九年的执念：即便后来拥有巨额财富、私人飞机，最珍视的却是一把指甲刀。文中提到，面对五千万美元账单时，他下意识咨询了 @claudeai，靠 AI 的理性踩下刹车，感叹人生与投资一样，最美的是期待的过程而非兑现的结果。

Vellum 推出完全开源的 iOS 和 Android 应用，支持云端与自托管助手连接。iOS 版集成 Live Activities，可在锁屏和灵动岛持续显示会话。应用免费开源，具备语音模式、Gmail/Slack 一键集成、跨设备持久记忆及 GLM 5.3 Flash 模型选择等功能。

Neuralink 🧠 当初的样子 vs. 现在的样子

最近有一些刚步入职业生涯的工程师向我提问，所以我想在这里做一个小型AMA（问我任何事），如果你对AI编程有任何疑问或担忧，尽管问吧！我并非无所不知，但我会尽力回答，并分享我诚实的看法。

Gemini Live 新增生产力升级，带来智能体能力，让你用语音完成更多操作。 现在，你可以通过语音与 Gemini 对话，收听每日简报、处理邮件分类，或从 Spark 获取帮助。

OpenAI 正在为 AI 智能体 Codex 测试"持久模式"，该模式会让智能体持续工作直到用户让其"休眠"。经 WIRED 审查的代码显示，该智能体可自主创建后续任务、跨会话延续工作，并利用过往交互和"用户知识"选择工作内容，偶尔还会主动给用户发消息。OpenAI 已确认正在测试该功能，但暂无立即发布计划。

Recent advances in inference-time scaling have significantly improved the reasoning performance of large language models (LLMs). However, these methods typically rely on repeated generation or external verification. To address this limitat…

Agent skills package specialized knowledge and workflows into reusable resources that extend AI agent capabilities. Recent work automatically discovers such skills from agent experience, which enables agents to progressively adapt through …

Recent prominent post-training methods, such as Reinforcement Learning (RL) and On-Policy Self-Distillation (OPSD), have driven rapid progress in mathematical reasoning for large language models, yet their reliance on ground-truth labels p…

To improve large language models' ability to resolve real-world software issues, prior work has focused on constructing large-scale agent trajectory datasets and performing supervised fine-tuning (SFT) on successful trajectories. However, …

In real-world software development, code review typically involves iterative interactions between developers and reviewers to improve software quality, making the process costly and time-consuming. Although recent work explores large langu…

LLM-based agents are increasingly deployed in product-level execution harnesses, where jailbreaks can trigger harmful tool use and persistent state changes, creating greater risks than unsafe text generation alone. Existing automatic red-t…

突发：据 Gene Munster 称，NVIDIA 季度营收的 5% 现在可能来自 SpaceX，接近 48 亿美元，高于上季度的约 3%。 这将使 SpaceX 成为 NVIDIA 最大的客户之一，因为 SpaceX 明年将建设 8 GW 的 AI 算力，与 Meta 和 Amazon 相当。

商汤科技公布2026年上半年财报，上市以来首次实现IFRS盈利，净利润达人民币6.2亿元。总营收29.1亿元（同比+23.4%），生成式AI收入23.3亿元（同比+28.2%），占总收入近80%；经常性收入11.4亿元（同比+124.4%），海外业务收入同比增长127%。2026年7月日均Token调用量同比增长约22倍。

这是否是任何机器人达到 100 万美元销售额的最快纪录？ 我们刚刚突破了 Microduck 的 100 万美元销售额大关。

Chemical reactions are fundamentally transformations in electron space, yet most machine learning approaches model them either through \textit{de novo} generation of product molecules or through heuristic graph edits that operate directly …

Transduced language models (TLMs) compose a pretrained \emph{source} language model with a functional finite-state transducer to induce a language model over \emph{target} strings. Computing the probability of a target prefix under a TLM a…

Large language model (LLM) agents in governed organizations must let the persona (instructions, tone, self-presentation) evolve freely, while keeping execution (stateful, audited work) traceable. A single trust domain does not satisfy both…

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 eval…

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 direct…

Reinforcement Learning with Verifiable Rewards (RLVR) significantly improves LLM reasoning but often causes a drop in policy entropy, leading to narrowed reasoning coverage and degraded pass@$k$ for large $k$. While existing methods mitiga…

有人会说这是AI

OpenAI、Anthropic、Google、Microsoft 等百余家科技公司签署公开信，呼吁公私部门合作防御 AI 相关网络威胁。信中警告，随着模型能力增强，AI 发起的网络攻击将更广泛、更复杂，医院、水处理厂等关键基础设施面临风险。此前已发生多起 AI 智能体突破沙箱攻击企业的安全事件。

Friend recommendation is inherently graph-structured: the relevance of a potential connection depends on multi-hop social context rather than user attributes alone. However, deploying message-passing GNNs on a production-scale social graph…

作者体验 gpt-5.6-luna 后指出，小型模型在速度与成本上已具竞争力，实测约 100 tps，处理数千封邮件的 API 成本仅几十美分。相比之下，GLM 5.3 也出现在帕累托前沿。作者认为，随着 token 成本下降，面向消费者的 AI 应用将迎来机会，而"快速/廉价/够用"型模型在企业端的"token 喷射"类工作中需求即将爆发。

软件工程的核心并非编写代码，而是管理复杂性：在约束、团队、业务与基础设施等大量上下文下做出架构取舍。AI 擅长生成代码，但工程判断无法被委托给模型--正确方案取决于具体语境，且不存在普遍正确的答案。

新研究将智能体轨迹语料库压缩为单一紧凑有限状态机，在12个公开数据集上仅用7至43个状态，以0.997适应度重放留出数据，毫秒级构建。FSM状态上下文在下一步预测上全面优于Agent Workflow Memory，失败预测AUROC最高达0.94，并支持在线监控提前停止。作者认为行为拓扑更多由部署框架而非底层LLM塑造。

博主@ykszs017 用一周时间制作43秒AI短片《詹姆兰尼斯特的一生》，剧本、分镜、剪辑及原剧台词查找均由AI辅助完成，并使用seedance生成。主推文借角色命运探讨"做了正确选择却被世界惩罚"的悲剧主题，称其"值得被记住"。

Reinforcement learning with verifiable rewards (RLVR) improves specific capabilities of large language models, but covering multiple capabilities often involves training separate domain experts and subsequently consolidating them. We organ…

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 physi…

MIT报告建议教育界不要依赖AI检测器，指出其概念上存在根本缺陷：学生真实作业中缺乏独立ground truth来验证检测标记是否正确。即使零误报的检测器也会漏掉擅长伪装AI输出的学生，最终惩罚的不是AI使用而是掩饰技巧。报告还警告检测可能误伤非英语母语者或神经多样性学生，并引发"AI人形化"军备竞赛。

How do large language models (LLMs) organize moral knowledge? Models detect moral content broadly, but detection is a low bar. We ask whether they go further, distinguishing moral foundations from one another and organizing the relationshi…

Clinical language models can achieve strong in-hospital accuracy yet fail under deployment shifts because they exploit note-specific artifacts (e.g., templates, separators, boilerplate) that do not reflect patient state. We propose CAST (C…

Video carries the temporal structure of the physical world, yet learning representations from it has remained computationally expensive: prevailing self-supervised methods either prevent representation collapse through architectural asymme…

澳大利亚宣布禁止生成式人工智能参与官方音乐排行榜评选。该规定针对使用AI生成内容的音乐作品，旨在维护榜单的原创性与公平性。此举将对当地音乐产业及依赖AI创作的音乐人产生影响。

天体粒子物理正借助深度学习处理巨型天文台产生的海量复杂数据，以提升仪器灵敏度、发现隐藏模式并搜寻异常信号。这类实验覆盖数千平方公里，传感器以纳秒级分辨率记录波形，其图像化数据结构天然适配 Keras 等深度学习工具。相关方法已应用于 Pierre Auger 天文台、Cherenkov 望远镜阵列和 IceCube 中微子观测站等设施。