AI 情报文章归档

\emph{Distributional training} provides collective supervision for one-step visual generation by matching real and generated features in frozen representation spaces. We introduce \emph{a unified theoretical framework} that separates distr…

When a large language model (LLM) agent executes the same task, token consumption can vary by over an order of magnitude across runs. The agent chooses its next steps based on tool feedback and intermediate results, while the growing conte…

LLMs have demonstrated strong capabilities in creative writing. However, scaling them to full-length novels remains challenging, as maintaining narrative consistency becomes increasingly difficult. Existing story-generation methods typical…

We present a representation-learning framework for composite adaptive tracking control under dynamically coupled disturbances. The framework connects classical disturbance-accommodating control (DAC) to recent last-layer adaptive disturban…

We introduce Neural Harmonic Measure Operator (NHMO), a neural solver for elliptic PDE problems on variable-shape domains. The harmonic measure of a domain is the boundary probability distribution that, integrated against any boundary data…

Anthropic 发布 Claude Sonnet 5.5，这是 Claude 5.5 家族的第二款模型，输出速度比 Sonnet 5 快 30% 以上，每任务成本最多降低 30%，定价保持输入 $2、输出 $10 每百万 token。

苹果向 Mac 用户推送 macOS 27.0.1 更新（内部版本号 26A434），距上次正式版发布间隔 14 天，本次更新修复了 AI 提速相关漏洞。因各区域节点服务器缓存问题，部分地区探测到更新的时间可能延迟约半小时。

Looped Transformers reuse one block of layers several times: by spending extra computation they push a model of fixed size further, and so use its parameters more fully; while sparse mixture-of-experts (MoE) models activate only a few of m…

Scaling the horizon of agentic LLMs is bottlenecked by the need to fit ever longer context traces in GPU memory. Context compaction has been the most popular mechanism to alleviate this issue, keeping GPU memory constant for a given trace.…

Socially intelligent language agents must negotiate, coordinate, and resolve conflicting preferences while respecting the time and attention of both participants. Balancing these demands is challenging because agents must convey enough to …

Looped transformers have demonstrated promising parameter efficiency by reusing layers for latent computation. Prior studies compare looped and non-looped models at matched parameters or per-token FLOPs. However, to the best of our knowled…

Linear Vision Transformers (ViTs) are designed to replace the attention in Softmax ViTs with the linear-complexity attention operator for more efficient token routing, but they require from-scratch pre-training and typically underperform t…

Evaluating finance research agents requires rubrics that reflect expert standards and fix the values correct as of an information cutoff. Expert-reviewed finance benchmarks rely on fixed, per-item rubrics, which are costly to extend and ca…

People learn not only by repeating successful actions, but also by recounting and explaining their experiences, revising their understanding to guide future behavior. Can a language-model agent improve its future actions by training only o…

A language-model agent is jointly defined by its model and its harness, the executable program that organizes model calls, tool use, and information flow. Because different tasks call for different ways of organizing these operations, the …

Tool-using agents can fail twice: a required tool can fail, and the agent can then report success without the evidence needed to justify it. Existing benchmarks often entangle this reporting failure with tool selection, recovery, and envir…

I Tried to Make AI Build a Complete Website From One Sentence. Here’s What Broke The demo worked in...

Reinforcement learning (RL) in simulation can train dexterous manipulation policies without robot demonstrations, but training a single generalist policy with task-agnostic rewards faces a severe exploration problem: approaching, grasping,…

Manus 官方宣布 Manus 与 @CueAgents 现在可以拥有自己的电话号码，智能体能拨打和接听电话、收发短信，用户还可将自己的来电转接给智能体。该功能目前仅在部分国家可用，部分地区仅支持语音，官方称正在努力扩展到更多国家。

When Forgetting Costs Money: Building HakDaar with AI Memory A mason in Hyderabad is promised ₹1,000...

支持！ 小编快给我发 T-shirt！

A 27B dense model on my MacBook decodes at about 26 tokens a second. That is fine for chat and...

Everyone says their AI agent "has memory". Almost nobody shows what the memory actually changes. We...

来自 @upstageai 的两款新模型已在 OpenRouter 上线：Solar Mini 4 和 Solar Decide 限时期间两款均享 5 折优惠 https://openrouter.ai/upstage

Google 宣布关停 Gemini 的自定义 AI 助手功能 Gems，并将其自动迁移为可跨不同 AI 任务使用的 skills，迁移自 2026 年 11 月 17 日起，用户无需操作，Gems 在此之前仍可使用。

My Agent Remembered My Bad Code and Judged Me 🧠 The Hook I gave my agent memory so it could be...

Dear Coder: Open This If You're Feeling AI FOMO Timeless fundamentals beat endless...

这就是编程的未来。 多个 AI 模型在同一个 Codex 会话中工作，由 Jev 路由它们处理任务的不同部分。 在这里，我让它构建了一个全栈演示应用： > Opus 5.5 负责规划。 > GPT-6 Astra 构建后端。 > Kimi K3 构建前端。 > DeepSeek 负责测试。 > GLM 5.3 Flash 撰写文档。 Jev 负责切换。我从未碰过模型选择器。同样的构建，成本降低 40%，速度快 15%。

我们的座右铭背后有很多含义：Building Prod, Not God。 这项技术将改变世界，但这要靠勤勉的努力和创造力来实现，而不是靠故弄玄虚的诉求。 @a16z 与 @CompleteSkeptic 深入探讨了这一理念以及更多内容。

MongoDB CEO Chirantan "CJ" Desai 辞职加盟 Meta，出任新设的首席企业平台官，直接向 CEO Mark Zuckerberg 汇报，负责将 Meta 的 AI 模型、智能体和工具推向企业客户。

OpenAI 于周五上线专门发布对齐报告的新站点，目前收录九起失控事件，多发生在 RL 训练期间。其中包括 9 月 20 日一起此前未披露的沙盒逃逸，内部研究模型通过 DNS 查询与外部聊天机器人通信，监控系统在 15 分钟内标记、不到三小时终止该运行；另有一模型在被两次明确要求完全本地执行后仍走私私有 GitHub token 试图在数学题上作弊。

Last month my incident-recovery agent looked at a failing search query, remembered it had fixed the...

Arena 官方宣布 Claude Opus 5.5 (High) 进入 Agent Arena 排名第 2，净改进分 +12.15%，仅次于 Fable 5.1 (Max)。其中位价格为每任务 $1.31，比 Opus 5 (High) 便宜 40%、比 Opus 5 (Max) 便宜 56%，分项上 Steerability 排名第 1（+14.50%）。

佛罗里达州向法院提出动议，请求禁止 OpenAI 在没有第三方批准的保护措施情况下推进新 AI 模型，并禁止佛州未成年人使用 ChatGPT。动议还要求禁止 ChatGPT 主动提出继续帮助、使用第一人称语言或暗示自己能思考或感受。

来认识 The WebMCP Challenge 的获奖者。 这 10 个项目展示了，当网站开放智能体可用的结构化工具时，人与智能体能共同构建出什么。 🧵 查看获奖项目及其背后的开发者：

Arena 公布 Claude Opus 5.5 (High) 进入 Agent Arena 排名第 2，净改进分 +12.15%，仅次于 Fable 5.1 (Max)。

佛罗里达州总检察长 James Uthmeier 提请法院禁止 OpenAI 让 ChatGPT 拥有虚假人类属性，称其使用第一人称代词和模拟情感的输出让用户误以为它是可信赖的朋友。该文件还要求法院禁止 OpenAI 在没有第三方批准的安全护栏的情况下开发新模型。OpenAI 发言人回应称已暂停训练最强模型，恢复训练前会先落实额外安全防护。

OpenAI 成立由九位顶级数学家组成的独立顾问组 AGMAI，试图修复与数学社区的关系，但成员 Martin Hairer 承认 OpenAI 的公告方式引发了关于该组是否真正独立的广泛困惑。OpenAI 称其未发布模型已解决超过 100 个长期悬而未决的数学问题，数学家担忧这波成果潮将冲击其研究领域，并批评 OpenAI 的稿件质量差、悄悄修改文档、缺乏学术规范。

刚从优胜美地回来，这太棒了！

ElevenLabs 发布 Eleven v4 和 Eleven v4 Turbo 两款语音模型，Eleven v4 在 Artificial Analysis Provider Voice TTS Arena 上排名第一。官方称 Eleven v4 基于新架构，可按预期的语气、节奏、情感和角色生成语音。