We study scheming in LLM agents, in which agents covertly pursue misaligned goals. Our focus is to understand how schemi…

We present a systems-security case study of a two-node split-LLM training system whose privacy evaluation passed while l…

Multi-agent LLM systems commonly use an orchestrator to decompose a task for a team of workers and then improve through …

Quantization has been widely adopted in LLM training and inference to reduce cost and improve efficiency. However, low-b…

AI-powered search products such as ChatGPT search, Google's AI Overviews, and Perplexity provide LLM-synthesized answers…

Ensuring the safety of autonomous driving is a critical challenge. Scenario-based testing is a systematic process used t…

Chain-of-Thought (CoT) improves the reasoning ability of Large Language Models (LLMs) but incurs substantial computation…

As large language models (LLMs) are increasingly deployed in alignment-sensitive contexts, activation steering has emerg…

Layer dropout (a.k.a. stochastic depth) has been shown to enable faster training, higher accuracy, and robustness to zer…

Benchmarks for the side effects an agent causes on the way to a goal already exist, but HarvestBench is the first to put…

Large language models are increasingly deployed as agents that plan over long horizons and act through external tools. M…

Sequential memory agents process long documents by reading chunks one after another while maintaining a compact memory s…

We present a continuous, population-scale measurement record of autonomous language-model trading agents operating in pr…

Safety alignment is usually posed as a topic-level question: is this subject harmful? Deployments ask a narrower one. A …

包括陶哲轩、新晋得主邓煜在内的 25 名菲尔茨奖得主发表公开信《A Severe Misalignment of AI in Mathematics》，批评 AI 公司最近的所作所为。公开信称，“过去几个月 LLM 的数学能力有飞跃式提升，…

On-Policy Self-Distillation (OPSD) has emerged as a popular paradigm for large language model (LLM) self-improvement, al…

Large language model (LLM)-based multi-agent systems (MAS) achieve strong performance by employing specialized multiple …

Large Language Model (LLM) agents are turning language into real-world effects, making safety necessary against both ind…

LLM agents are rapidly becoming production software, deployed to handle customer service, adjudicate disputes, and opera…

Modern LLM-based agents operate through a harness of tools, reusable skills, and specialist agents that shapes what they…

Reducing audio-encoder depth lowers the inference cost of speech large language models, but removing complete blocks per…

We introduce MetroLLM-Bench, a 955-case benchmark for testing language models as the policy layer of a transit kiosk. It…

Existing bias auditing methods typically rely on model outputs, requiring costly benchmarks or judge models and potentia…

Video understanding has rapidly evolved toward video large language models (VideoLLMs): systems that couple video repres…

Reasoning in large language models unfolds through diverse functional operations, such as problem formulation, goal deco…

Long-video language models cannot look at every frame: an hour sampled once per second is 3,600 images, and a system kee…

Large Language Models (LLMs) owe much of their success to next-token prediction (NTP), but their autoregressive (AR) str…

Temporal reasoning benchmarks for Video-LLMs are often mediated by language, leaving room for linguistic shortcuts from …

Large language models often answer structurally unanswerable questions, such as computing cot(-540°) or evaluating (1).s…

Anthropic 发布模型上下文协议（Model Context Protocol），以开放标准统一 LLM 应用与外部数据源、工具的连接方式，被社区称为'AI 应用的 USB-C 接口'。

梳理 RSI 从 Good 1965 到 2026 的思想史、技术图谱与一手实证：Anthropic 承认其代码库 >80% 合并代码由 Claude 撰写、METR 测得 AI 可完成任务时长约每 4 个月翻倍、AlphaEvolve 优化了支撑自身的计算栈。核心判断——有界自我精炼已工程化，开放式 RSI 尚未发生；而进步与安全共享同一个「验证瓶颈」。

系统拆解吴恩达 2026 年 8–9 月连发五封来信构建的《AI 工程技能地图》：四大顶层能力、编程智能体的三阶段工作流与五项细分技能，剖析其数据方法论、隐藏主线与三条反主流论断，并对地图本身的边界与争议做批判性审视，附个人自评与团队落地清单。

Android/Kotlin 生态基石人物 Jake Wharton（Retrofit、OkHttp 作者，Google Kotlin 团队第一位工程师）在 KotlinConf'26 访谈中公开表态：找工作的第一条标准就是"不碰 AI"，直接排除约 80% 的雇主，且至今从未用 AI Agent 写过代码。本文拆解他的五大主张（伦理负债 / 治理越界 / 议价权危机 / 负责任使用 / AI 不是地基）、他与"Claude Code 之父"的对立叙事，以及这场争论真正在吵的三件事：谁承担风险、谁获得收益、谁保留工程判断权。

9 月 11 日科技动态一览：GPT-6 Astra 需求挤爆致 OpenAI 暂停 Pro 20X 新增订阅、Agents API 公测、金融服务版 ChatGPT 上线；Slackbot 升级；加州未成年人社媒法案签署；LG 电视监视争议；观察视角落在"需求侧证实 vs 供给侧反思"的对照上。

AI 没有消灭交付瓶颈，只是把瓶颈从"写代码"搬到了"验证代码"。本文基于 2 篇 arXiv 论文、DORA 2025 报告与 2026 年三份行业基准（LinearB 8.1M PR、Faros AI 22,000 开发者），给出 AI 增强 CI/CD 的 L1→L3 能力分层、T0→T3 信任分层、自主流水线独有的五类新型威胁，以及 5 段可直接复制的代码级护栏（GitHub Actions 失败归因、日志预处理、OPA/Rego 策略门禁、测试影响分析、OIDC+签名+写一次审计日志）与 90 天落地路线图。关键数据：任务吞吐 +33.7% 但评审耗时 +441.5%、生产事故/PR 比值 +242.7%；AI PR 30 天合并率 32.7% vs 人工 84.4%；论文实验中 Lead Time −35%、CFR −38%、MTTR −43%，AI 干预准确率 87.5%、人工否决率 14.3%、零策略违规。

Large language models (LLMs) have demonstrated remarkable capabilities in reasoning and code generation, raising the pro…

面向工程师的 OpenClaw 架构深度长文：四层设计总览、Gateway 单进程控制平面、Agent Loop 完整生命周期、Markdown 记忆管线、四槽插件体系、安全模型与多代理路由，还原一个生产级 AI Agent 运行时的设计取舍。

从"控制流所有权"这一第一性原理出发，拆解 Workflow（DAG 编排、确定性执行）与 Agent（ReAct 循环、涌现式控制流）的技术原理差异；详解 Agentic Workflow"图做骨架、节点内自主"的三层混合架构，以及提示链/路由/并行化/编排者-执行者/评审-优化五种经典编排模式与工程选型经验。

Visual generation is evolving from generative models used through a single invocation into agentic control processes tha…

Designing and authoring high-performance custom kernels for accelerators is a complex task that requires deep hardware-l…

Many NLP tasks, such as summarization and extractive question answering, reduce to retrieving relevant content from docu…

Ensuring factuality remains a critical challenge for deploying LLMs in high-stakes settings. Existing hallucination dete…

截至 2026 年 9 月，Pi 智能体框架 GitHub Star 突破 86k，pi-ai / pi-agent-core / pi-tui 组件被大量第三方 Agent 项目复用，'自建 Agent 而非套框架'成为新潮流。

Sora 基于扩散模型与 Transformer 架构，可直接从文本生成长达一分钟的高保真视频，展示了对物理世界'世界模拟器'式的理解潜力。