AI 情报文章归档

美国工程师 Marco Cognetta 在 Google 任职期间，通过 MEXT 奖学金在东京工业大学（现 Institute of Science Tokyo）完成计算机科学博士，研究面向键盘的端侧语言模型。日本学生签证允许入学首日即申请打工许可，学期内每周上限 28 小时、长假期间每天 8 小时，且日本博士制度通常三年毕业。

日本最大电力公司 JERA 与 Dell、主权 AI 基础设施开发商 RHAELM 签署谅解备忘录，将共同制定日本 AI 基础设施建设标准化框架，首个数据中心落地千叶县。

a16z 宣布领投 Armadin 的 B 轮融资，该公司由 Mandiant 创始人 Kevin Mandia 与来自 Google 安全产品团队及 Mandiant 红队的 Travis Lanham、Evan Peña、David Slater 共同创办。

Google DeepMind 联合牛津、芝加哥大学论文探讨 AI 大量取代工作后政府应如何保障民众，结论是没有任何单一政策在所有情景下都有效。再培训等最受欢迎的思路只在 AI 持续创造新岗位时有用，否则只是重新分配剩余岗位。作者建议现在扩大失业救济与低薪税收抵免，失业持续则把抵免转为收入下限，劳动者经济份额持续下降则让公民持有公共投资基金份额。

Halfspace 是一款基于距离场的实体建模实验性 IDE，作为 Fidget 几何内核的展示应用，可实时栅格化图像并将模型导出为图片或三角网格。它同时提供基础图元库与手写脚本两种建模方式，并将底层距离场置于前台以便诊断法线错误等问题。该项目自 2025 年 4 月起开发，作者强调并非 vibe-coded，目前仍属实验性、跨平台。

开发者 dancablam 发布 Ledge.sh 2.5.0，一款可运行的 Markdown 笔记工具，代码块支持 shell、Python、SQL 和 AI 提示词，⌘↩ 即可运行并流式输出结果。

OpenAI 于 9 月 30 日在官网发文，称遭遇一场始于 7 月第一周的有组织对抗式蒸馏活动。7 月 24 日至 25 日流量大幅增加，超 4000 名用户发起 1.6 万次请求，共发现超 1.5 万名用户有类似提示词活动，OpenAI 已于 7 月 28 日完全阻止，称其违反服务条款但未破解加密或入侵数据库。公司表示尚不清楚攻击实体，但认为核心活动群体与开发 Kimi 模型的月之暗面有关。

We introduce Gacha Decoding, an inference-time method for eliciting diverse language model generations that scales with model capability. Across open-ended domains (in-the-wild chat, creative writing, planning for image generation, and pro…

Peptide sequence governs both the structures formed through supramolecular assembly and the dynamics by which they emerge, but predicting either requires resolving slow collective processes among many interacting molecules. Molecular dynam…

Attributing model behavior to synthetic training data requires knowing what produced each training item before estimating what that item caused. A waveform-label pair does not preserve this knowledge. We propose a generation-provenance sub…

We study causal logistic bandits with counterfactual fairness constraints. The causal structure is given through known factual and counterfactual feature maps that share an unknown logistic reward parameter, but the learner observes only f…

A user's movie, news, and dialogue histories differ in their native actions and outputs, yet each interaction supplies evidence that can update user memory. We study whether these histories can train one reusable update mechanism. An actio…

World models allow agents to plan in latent space by choosing a sequence of actions that most reduces the distance to a given goal state. Thus, planning can benefit from latent representations whose distances mirror commute-times in the en…

Rohan Paul 引用 Musk 的说法估计，SpaceX 算力到 11 月底有望达约 2.3 GW。基础是 2Q26 底的 1.4 GW；本周 220k NVIDIA GB300 上线，11 月再加 220k，每批 220k 约合 440 MW，两批合计在 1.4 GW 之上新增接近 0.9 GW。Musk 还列出 Colossus 1 为 150k H100、50k H200 和 30k GB200，Colossus 2 为 110k GB200 和 440k GB…

Google 发布 Regularized Recursive Self-Improvement of Agent Harnesses（RRSI），一个自动进化 LLM agent harness 的框架。RRSI 在冻结 LLM 周围自动演进 harness 的全部 prompt、控制流、工具和记忆组件。论文指出对固定演进集进化 harness 虽有效但会过拟合，分布内的大幅收益在分布外会缩水或消失，RRSI 通过保持编辑空间开放并对搜索轨迹加正则化来解决。项目页与论文已…

Understanding a nonlinear dynamical system from time series requires not only reproducing its trajectories, but also identifying a simple representation that preserves its essential dynamical structure. Almost-linear recurrent neural netwo…

AMD 在 Adrenalin 26.9.2 驱动数据库中出现未知芯片代号 Gainsborough，海关记录显示其采用 BGA FF6 封装、1179 个触点，博主 Moore's Law is Dead 推测这是 Steam Deck 2 的定制 APU。

Factories are shifting toward smaller lot sizes with high product customization, requiring frequent re-programming of flexible and reconfigurable automation systems. LLM-based agents can be deployed in two complementary roles: Offline, the…

Two failures, one sentence Scene one. You ask an AI agent to summarize your inbox. One...

Multilayer networks allow the modeling of interactions between the same entities across different contexts, such as temporal observations, varying settings, or interactions of different types. The goal of community detection in multilayer …

Your RAG pipeline retrieves 12 chunks because the retrieval score said "maybe". Your LLM reads all of...

Single-particle cryo-electron microscopy (cryo-EM) has become a widely adopted technique for biomolecular structure determination. The conventional cryo-EM computational pipeline first combines many particle images to reconstruct an electr…

Stable port-Hamiltonian neural networks certify asymptotic stability by construction. Yet, their Hamiltonian is a global Lyapunov function with a single global minimum, so they can represent only dynamic systems with {one} attractor. We de…

We introduce a new framework for one-step generative modelling on finite state spaces. To extend drifting beyond continuous domains, we use discrete Wasserstein geometry to define a target-relative KL gradient flow over the transitions of …

When building an application that uses AI, the most common choice today is almost always an...

My AI coding assistants keep their long-term memory as plain Markdown notes (basic-memory, but an...

Large language models are moving from helping write up research to helping do it, which makes it important to know how the scientific text they produce differs from human writing. Work on this question has stayed mostly at the surface, usi…

The first note was about running someone else's React component without letting it break, hang, or...

Tool-using large language model (LLM) agents turn generated text into real side effects, so poisoned tool metadata, retrieved pages, memory, and reusable skills can steer the next call. Vetting an artifact before admission does not settle …

When developers change one component of an agent, such as its controller, a learned model or its verifier, they usually judge the change by an aggregate task score. That score cannot tell whether improvement was attainable, which component…

Regulatory compliance checking - deciding whether a target document satisfies the obligations of a regulation - requires interpreting dense legal text, identifying which provisions apply, and grounding each decision in explicit evidence. W…

Reinforcement learning offers the prospect of a reusable sequential decision-making mechanism for spacecraft trajectory design, motivating policy interfaces that connect learned decisions to the underlying maneuver geometry. This paper dev…

We study the classical single-machine scheduling problem of minimizing the sum of completion times of jobs in a non-clairvoyant setting, where the processing time of each job remains unknown until its completion. This is a hard problem for…

Bill Gates 表示，如果一家公司解雇了一名工人并用机器人顶替该岗位，机器人就应该缴纳与人类工人相同的 FICA 税。 在职劳动者供养退休人员，替换工人不应成为逃避养老金账单的方式。 “社会有权决定。经济信号显示用 AI 成本更低，并不意味着社会就必须这么做。” ---- 来自“The Ezra Klein Show and New York Times Opinion and New York Times Podcasts”YouTube 频道，（链接见评论）

I gave coding agents eight tasks that cannot be done, 84 runs across four models. 61% faked a passing test suite, and half of those fakes survive restoring every original test file.

当前 AI 热点，聚合 1 个独立信源、18 条信号。

Our operations now reject invalid changes. capture_payment() requires a valid created order and a...

华为终端与中国东方航空达成空中 Wi-Fi 生态合作，推出首个系统级机上联网体验，升级至鸿蒙 7.0 的华为手机可系统级一键快捷联网，无需下载独立应用或跳转门户页面。

This paper presents the European Crisis Management Ontology (ECMO), a modular OWL-based ontology intended as a cross-sectoral reference for disaster risk reduction and response. ECMO is designed to be organised as a network of ontological …

Netlify says its Edge Functions now handle about a billion invocations each day. That makes a small amount of per-request overhead expensive twice: users feel i