AI 情报文章归档

Meta Muse 工程副总裁 David Singleton 宣布 Muse 拥有自己的邮箱地址，可代用户收发邮件。该功能已向美国和加拿大的付费订阅用户开放，并将很快对所有用户开放，用户可决定它联系的对象，Muse 在用户批准下通过自己的收件箱工作。

Amazon 发布了一款安防摄像头无人机，警报触发时会自动飞往家中预设位置，Ring 门传感器或警报被触发后，无人机或地面机器人会自行前往，用户打开警报时视频可能已在直播。Ring 无人机停靠时摄像头关闭、仅在飞行时录制，螺旋桨故意设计得响亮，两款机器人的视频默认采用 Ring 最强加密。

OpenRouter 官方宣布正式并入 Stripe。OpenRouter 称有超过 1000 万开发者和公司使用其平台，它为多模型、多 provider 的 AI 运行提供基础设施，包括访问新模型、故障期间保持可用，以及为团队执行支出和数据策略。

你委派的智能体需要一个人们能联系到的地址，以及一台能在任务之间保持状态的机器。 @inkbox_ai 已经为智能体提供了 email、电话、iMessage、WhatsApp 和 A2A。从今天起，它还负责运行智能体，基于 Hermes 或 Codex，带调度功能和持久化计算机。 祝贺他们完成 530 万美元融资，让 AI 智能体拥有可运行的计算机

Cursor 宣布 Gemini 4 Carbon 现已在 Cursor 中可用。官方称在高努力（high effort）档位下，该模型以 60.4% 的成绩位居 CursorBench 榜首，配图显示其成本为 $9.22。作者推荐试用该模型。

用 Reframe 将现有素材适配到新的宽高比，无需重拍。 智能扩展并重新构图场景，同时保留原始主体与视觉一致性。

fal 宣布 LTX-2.5 系列模型上线，提供 Alpha Gen、Reframe、Upscale 和 SDR to HDR 四个入口链接。模型由 Lightricks 提供，用户可通过 fal.ai 平台直接体验。

LTX 的全新 VFX 能力现已在 fal 上线。 用四款强大工具生成并编辑达到制作水准的视频： - AlphaGen：生成边缘细节精细的透明蒙版，最高支持 4K - SDR → HDR：将标准视频转换为 HDR，扩展动态范围 - Reframe：将视频适配到新的宽高比 - Upscale：将视频分辨率提升至最高 4K 或 8K 在 fal 上使用 LTX 构建。

Anthropic 再次修订 IPO 申请文件 S-1（Amendment No. 2），宣布已越过其 Responsible Scaling Policy 设定的阈值，即 AI 正大幅加快 AI 研究速度、需更强安全检查。这一判断基于其对未来模型的预期而非现有模型的表现，一周前的文件还只说"可能即将"越过。文件同时将 IPO 发行价定为每股 $70 至 $78。

英伟达 CEO 黄仁勋承认与一名疑似中国间谍的女性有染，事件可能冲击整个美国 AI 产业并危及国家安全。专家向 The Verge 表示，若中国借此获得英伟达芯片，美国在 AI 上的领先优势"可能在数周或数月内消失"；与这名疑似间谍有关联的公司据称已促成至少 70 亿美元的英伟达芯片销往中国。消息曝光后英伟达股价下跌。

PSI 实验室的 David Turturean 借助 Anthropic Fable 5.2 模型早期访问权限解决了存在 18 年的量子计算问题 Aaronson-Ambainis 猜想，证明经 Lean 自动形式化验证。

Google 宣布 Gemini Intelligence 扩展到更多设备，包括 motorola signature 27（美国 11 月 19 日开启预售）和通过 Pixel Drop 获得新功能的 Pixel 10 系列。

Foundation models pretrained on large natural-image collections are increasingly adapted for medical image segmentation, in the expectation that they will replace models trained for each task. Most comparisons, however, use data from the t…

DeepLearning (DL)-based Computer-Aided Diagnostic (CAD) systems have shown great promise in supporting clinicians, but their performance heavily depends on the availability of large, well-annotated datasets. In medical lesion detection, th…

We present TorchCurves, a PyTorch library that treats vector-valued parametric curves as learnable primitives. Its central operation maps batches of scalar inputs to points on curves represented in B-spline or Legendre polynomial bases, wi…

Reinforcement learning with verifiable rewards (RLVR) produces rich training experience, yet outcome rewards reduce each attempt to a single number and leave most of this information unused. Training-time scaffolds, such as strategies or h…

LLM agents are increasingly used for open-ended, data-driven scientific discovery, yet most of them evaluate hypotheses pointwise, testing an empirical claim $h$ only against its complement $\neg h$. When evidence rejects $h$, the agent le…

Predicting how a ligand binds to its target is central to small-molecule drug discovery. Modern cofolding models can generate diverse candidate poses, but reliably ranking these candidates remains challenging. We introduce Pyrite, a pairwi…

Latent Space 评测面向编码的开源权重模型 Naive-N0.5-Flash，该模型由 NaiveAI 在小米 MiMo-V2.5（310B 参数 MoE、15B 激活、48T tokens）基础上继续训练约 3.25T tokens 并适配 DeepSeek Sparse Attention。

Connectome-constrained models are often used to argue that a measured wiring diagram outperforms shuffled or rewired wiring, and hence that structure shapes function. Such comparisons can confound wiring with how the models are tuned: a co…

Agent harnesses have emerged as a promising mechanism for continual learning, alongside test-time parametric learning. However, the current benchmarks have several limitations: they generally focus on a single applied domain, serve each in…

Foundation models trained on large unlabeled corpora develop emergent capabilities, tasks that go beyond the training objective. We extend this paradigm to multimodal biology with Bonbon, a foundation model for molecular interactions train…

Last week I asked a teammate, "Is the build green?" He replied with four paragraphs. There was...

Arena 宣布 MiniMax 的 H3-Image 图像生成模型已上线 Arena，用户可在 Battle Mode 中用创意提示词测试并投票，评分将陆续公布。据 MiniMax 引用内容，该模型支持 1600 万像素分辨率、10 秒内生成 4K 图像、最多 16 张参考图并开放权重，现已可通过其 API、MiniMax Code 和 Hailuo AI 使用。

Claude Devs 宣布 MCP Apps 相关能力以公开预览形式推出，实现细节可能随与 MCP Apps 社区的标准化工作而调整。官方提供了入门文档：https://claude.com/docs/connectors/building/mcp-apps/getting-started。

Anthropic 宣布在 MCP Apps 基础上进一步开放能力，开发者可将完整网站托管在 Claude 内，并让 Claude 通过工具对屏幕上的内容执行操作。MCP Apps 于今年早些时候推出，用于在 AI 应用中呈现交互式视图。

Claude 官方账号宣布 Claude 现在支持更多 MCP Apps 扩展能力，包括应用工具和嵌入站点，完整应用可以在 Claude 内运行，并给出开发入门文档链接。其引用的 Claude 官方内容补充称，交互式 connectors 可在对话旁打开 Adobe、Atlassian 和 Box 的完整工具版本，可自行修改或让 Claude 协助修改，正在 web 和桌面端推出。

Deploying LLM-based analytics agents in enterprise settings requires evaluation frameworks that can reliably detect failures across complex, multi-tool workflows. We present a three-phase comparative study of three evaluation frameworks, e…

OpenAI 于周二推出测试版 Business Hub，企业可创建并验证业务资料、发布业务智能体回答用户提问。Business Hub 还提供类似 Google Analytics 的有限洞察数据，包括资料浏览量、品牌提及和网站点击等指标，此前 OpenAI 在 3 月放弃了收取佣金的电商结账方案。

Business process automation runs the same Standard Operating Procedure (SOP) again and again on new inputs. Large Language Model (LLM) agents can follow an SOP written in natural language, but they call an LLM to choose every step of every…

Time series analytics has been widely applied across various domains. While existing time series foundation models (TSFMs) have demonstrated promising results, they still face two major challenges: 1) Most TSFMs mainly rely on stacking Tra…

Research ideation is iterative: researchers refine their ideas step by step based on feedback. Yet existing evaluations obscure this search process, making it difficult to understand how LLMs explore the space of research ideas and what li…

Static scanners read npm packages without running them, so they miss the postinstall that only fires at install time. I gave my AI agent one tool that runs the install in a disposable AWS CodeBuild sandbox and catches the curl | bash from …

据 Bloomberg 报道，Anthropic 启动 IPO 路演，拟发行最多 8.25 亿股 Class A 普通股，定价 70 至 78 美元，估值最高约 1.4 万亿美元，股票将以代码 ANT 在纳斯达克上市，最快下周四或周五开始交易。按区间上限募资约 644 亿美元，其中约 625 亿美元来自 Anthropic 自身的 8.01351 亿股新股，现有股东仅出售 2364.8646 万股，所得几乎全部归公司；此前传闻估值曾高达 2 万亿美元，下调发生在科技股抛售与…

Neural networks are increasingly studied not merely as predictive systems, but as complex systems whose internal computations are themselves objects of scientific inquiry. Recent progress in mechanistic interpretability (MI) has made it po…

In 2017, eight researchers published a paper with the cockiest title in computer science: Attention...

Nvidia 发布基础模型 GR00T 2，采用世界动作模型（WAM）架构，同时预测下一步动作和未来短片段视频，用超过一千块 GB300 芯片和约 40 万小时头戴相机人类动作视频训练，可将人类动作迁移到机械臂。

In this paper, we present 1020 Group's submission to the WMT26 shared task on automated quality evaluation. Our system, Pterodactyl, was engineered specifically for subtasks 2 and 3, that is, segment-level quality score prediction and the …

Anthropic 启动 IPO 路演，拟以每股 70 至 78 美元发行最多 8.25 亿股 Class A 普通股，估值最高约 1.4 万亿美元，股票代码 ANT，最快下周四或周五在 Nasdaq 开始交易。

Text-to-SQL systems over large pooled catalogs must first retrieve the tables a question needs. We compare against the published table recall of RASL, which uses Claude 3.5 models to extract keywords and select tables, on pooled BIRD, Spid…