AI 情报文章归档

相较于 Haiku 4.5，Haiku 5.5 在我们几乎所有的对齐评估中都展现出显著改进，行为失准的情况大幅减少。

Anthropic 发布 Claude Haiku 5.5，对 10 万 token 以下的任务提供更低定价。官方称此类任务约占上一代 Haiku 模型请求量的九成；定价表显示 10 万 token 以下输入为每百万 token $0.10、输出 $0.50、缓存读取 $0.01，超过 100k 后输入 $0.50、输出 $2.50，对比 Haiku 4.5 输入 $1.00、输出 $5.00。

Anthropic 发布 Claude Haiku 5.5，定位高吞吐、成本敏感任务，可靠处理摘要、分类等重复性工作，并可作为 subagent 配合 Claude Opus 5.5 和 Sonnet 5.5 做编码工作。官方称其速度足以支持实时客户支持和浏览器使用。

Haiku 5.5 在编程、计算机使用和知识工作方面相比 Haiku 4.5 有显著提升。

Anthropic 发布 Claude Haiku 5.5，称其为旗下最便宜、最快、能力最强的小模型。官方表示其平均运行成本比 Claude Haiku 4.5 低约 75%。

OpenAI 发布面向更广泛用户的 GPT-6，并随 GPT-6 在 ChatGPT 中引入 Intelligent UI，可生成图形、按钮、表单、图表和可交互组件来回答问题。

OpenAI 随 GPT-6 推出名为 Intelligent UI 的新界面，在对话中加入可点击按钮、定制计算器、交互图表和可编辑图形等互动视觉元素。该功能随 GPT-6 面向 Pro、Plus、Business 和 Enterprise 用户全球推送，免费版和 Go 档用户周四可用，用户可自行调节视觉内容的多寡。

As we move through the world and carry out everyday tasks, we encounter objects that may become relevant only later. We are capable of recalling where we left something or what was inside a container, even without knowing we would need it …

Unsiloed 发布 Higher-Order Extraction，可从页面原始数值生成确定性代码，计算出文档中并未直接印出的数字，每个结果都能追溯到页面。若文档某处恰好印有答案，系统会与自身结果比对：一致则置信度提升，不一致则置信度降为零。

Modern language models undergo reinforcement learning with verifiable rewards (RLVR) on top of already-trained checkpoints. A key promise of RLVR is the discovery of new reasoning strategies. In principle, a model can sample novel ideas ab…

Conditional memory architectures such as DeepSeek Engram use input n-grams to look up learned embeddings, expanding the capacity of large language models (LLMs) with limited additional computation. Beyond model scaling, this architecture h…

CZ Biohub 牵头一项 18 亿美元计划，训练 AI 模型预测细胞行为以加速药物研发，其中 4 月已承诺 5 亿美元用于五年期 Virtual Biology Initiative。

Real-time robot control demands enough visual history to infer motion and task progress, but processing that history can delay action. We present Long-WAM, a model-system framework for scaling the context of causal world-action models unde…

Heavy-tailed noise has been widely observed in modern machine learning, motivating the use of methods like gradient clipping and normalization. While these methods are well understood in centralized settings, much less is known in decentra…

Smith 因用约 1 万个机器人账号和 AI 生成歌曲伪造流媒体播放量骗取版税，被判 18 个月监禁，并须没收 8,091,843.64 美元非法所得。DOJ 称这是此类起诉中的第一案，指出其行为摊薄了平台版税池、损害真实艺术家收益；其律师以当事人未实际持有 800 万美元、已获唱片公司工作为由争取缓刑，DOJ 则以持续欺骗手段和说谎为由主张从重判刑以示震慑。

Vision-language-action models (VLAs) are strikingly sensitive to instruction phrasing and do not inherit the language robustness of the vision-language models they are built on. A one-word edit can move success by tens of points: $π_{0.5}$…

GNN-to-MLP distillation aims to retain the predictive accuracy of a message-passing teacher while deploying a graph-free MLP at inference. Existing methods mainly transfer node-wise predictions or use confidence-based reweighting, but they…

Machine unlearning asks for a deletion algorithm whose output is close to retraining from scratch without the selected forget examples. In this work, we study forget-only unlearning, where the deletion algorithm receives only the trained m…

Latent world models have shown a remarkable ability to predict future states and to plan in the real world. In practice, however, we lack a principled way to estimate how their capabilities scale with model size, data, and compute, an open…

Language-model agents are increasingly asked to carry out open-ended scientific research, yet their results are usually graded against a known answer, a rubric, or a language-model reviewer, none of which can tell whether a new scientific …

llama.cpp 可通过 ggml RPC 后端在异构设备间分布式推理，作者 Georgi Gerganov 表示这是高级设置，后续会让普通用户更容易使用。被引用的实测称，MiMo 2.6 Flash 的原生 mxfp4 权重可跨 RTX 6000 GPU 与 M5 笔记本以 40 tokens/sec 通过 10 GbE 运行，llama.cpp 开箱即支持。

Prompted embedding models have recently received increasing attention, particularly for retrieval, where detailed retrieval instructions are provided as part of the retrieval prompt. Several new datasets and studies have examined this sett…

Large language models are increasingly applied to tasks grounded in long, heterogeneous information sources. Conventional Retrieval-Augmented Generation (RAG) relies on fixed similarity-based retrieval, while agentic variants adapt queries…

Many of the questions now put to large language models have no correct answer to score against: what a policy is worth, which option a user should choose, how to weigh competing values. Stated-preference economics has faced this problem fo…

Google 将 SynthID Detector 向公众开放，任何人都可检测图片、视频或音频是否由 Gemini、Veo、Lyria 等 Google AI 或其合作方模型生成，支持 JPG、PNG、MP4 和 MP3 格式。

Online learning is an attractive framework in many domains because it permits well-defined learning even when data are dependent or chosen adversarially. This generality, however, comes at a steep price, introducing significant statistical…

Docker 发布 docker-agent，一个 Docker CLI 插件，用 YAML 声明式配置构建、运行和分享 AI 智能体，无需写代码，命令为 docker agent。

Robot foundation models provide strong visuomotor control, yet their performance can degrade when object positions or task instructions change. Further improvements often require post-training on substantial robot data, which can be costly…

微软在旧金山发布会上展示了全新的 Windows 搜索体验，借助 AI 让用户以自然语言查找信息、文件和应用，不再依赖关键词匹配。新体验强调理解用户意图，结合文件内容、应用信息和系统上下文返回更相关结果，并有望从“找到项目”进一步发展为帮助用户完成后续操作。该变化与微软推进的 Copilot、AI 智能体和“混合智能”理念相呼应，人类用户仍掌握最终判断和控制权。

Small tabular datasets with expert-designed spectral features are the norm in operational Earth observation, and the networks applied to them are typically hand-designed. We revisit one such published model -- a Sentinel-2 algal bloom clas…

I wrote a task for Terminal-Bench, a public benchmark of hard command-line tasks for AI agents. The...

微软面向开发者的 AI 迷你主机 Surface RTX Spark Dev Box 开启预售，11 月发货，售价 5,999 美元。

微软在 Surface 新品发布会上宣布，Meta 的 Muse 智能体将以原生应用形式登陆 Windows 平台，并集成 MXC SDK。Perplexity 的“便携电脑”工具可在端侧执行复杂 AI 任务，无需消耗云积分；OpenClaw 也将推出原生 Windows 网关。Windows 系统还将新增“入门指南”，帮助用户配置和运行 AI 智能体。

GitHub 发布与 Microsoft Applied Sciences 合建的 ModernBERT 微调分类器，用于将 push protection 扩展到非结构化密钥，单批评估不到 2ms，预计可阻止的密钥数量翻倍以上。

We study the best arm identification problem in a stochastic environment with a novel form of adversarial perturbations, which we coin Shifting Means. While classically the mean rewards of the $K$ arms are stable in time, in Shifting Means…

Our MCP server's tool descriptions were full of instructions to the model. Anthropic's directory review doesn't allow that. Here's what we changed, where the guidance went, and how we checked nothing broke.

微软在旧金山发布会上提出“混合智能”（hybrid intelligence）理念，让 Windows 11 的 AI 按任务类型、设备性能和安全需求在本地与云端之间协同：隐私敏感、需快速响应的任务更多依靠本地处理，复杂模型和高算力需求交由云端完成。

Sampling from a softmax distribution is a fundamental operation in machine learning, but its linear complexity in the number of items makes exact sampling impractical at scale. Two-level softmax (2LS) sampling is a popular alternative enab…

How can we predict which base checkpoint is worth an expensive round of agentic post-training? End-to-end pass@$K$ tests whether successful behavior already appears in a base model's distribution, but it is a poor fit for agentic coding: m…

grok @bot 0.68.1 发布！ - Bot 可以和你一起制作幻灯片，并以 PowerPoint 或 Google Slides 形式交付 - 直接从草稿卡片发送格式化邮件 - 你在 1:1 聊天中的消息会采用该 Bot 的颜色 - 更快的 computer use，现在支持 1920x1200 屏幕 https://x.ai/changelog/bot