AI 情报文章归档
浏览 AI圈报 已保存的 29154 条 AI 情报文章,覆盖模型、产品、行业、论文、教程与观点方法六大分类。
已收录文章29154
正式分类6
精选内容4420
全部文章
第 193 / 729 页 · 每页 40 条
论文研究普通
Low-Discrepancy Dither for Quantized Recurrent State Caches
Mamba-style and hybrid language models compress their past into a fixed-size recurrent state that is rewritten at every generated token. Storing this state in low precision saves memory bandwidth, but every rounding error is fed back into …
信息来源:arXiv
论文研究普通
Fiber-Resolved Microstructure Quantification from Multi-Shell Diffusion MRI using Detection Transformers
Fiber orientation and compartmental microstructure are central to the characterization of white matter tissue in diffusion MRI, yet existing methods either resolve fiber orientations without quantifying microstructure, or quantify microstr…
信息来源:arXiv
论文研究普通
MEND: Label-Free Detection, Localisation, and Correction of Latent Hallucination in World Models
World Models are appearing as the next major frontier in computer vision. However, their robustness is currently largely unexplored. We identify the phenomenon of hallucination in latent World Models: given a state and an action, the predi…
信息来源:arXiv
行业动态普通
LG 与微软合作研发智能家居 AI 语音助手,搭载 Voice Live 应用于 ThinQ ON
LG 电子在首尔微软行业峰会上展示与微软合作研发的智能家居 AI 语音助手,基于语音到语音智能体技术并融合微软 Voice Live,支持用户插话提问时立即中断原回答并切换对话。该助手将应用于智能家居中枢 ThinQ ON,LG 计划年内推出搭载 Voice Live 的版本并为现有用户更新,场景还可扩展至酒店、办公室、零售门店等 B2B 空间。
信息来源:IT之家(RSS)
论文研究普通
Whitening Improves Robustness to Spurious Correlations in Linear Probes
Deep neural networks tend to rely on simple features that may be spurious and thus fail to generalize. We study this problem in the setting of linear probes, where a (generalized) linear model is fitted on the representations of a (pretrai…
信息来源:arXiv
行业动态普通
英伟达测试无玻纤HC覆铜板用于Rubin Ultra
郭明錤供应链调查显示,英伟达已开始测试以碳氢树脂为主的无玻纤HC覆铜板(CCL),替代此前测试的无玻纤PTFE基CCL,搭配无玻纤HC PP,用于Rubin Ultra NVL576交换机托盘PCB,该八机架互联方案预计2H27量产。
信息来源:X:郭明錤 Ming-Chi Kuo (@mingchikuo) · NVIDIA
行业动态普通
Nvidia 测试 HC 无布 CCL 用于 Rubin Ultra 交换机托盘
郭明錤最新产业调查显示,Nvidia 开始测试以碳氢树脂(HC)为主的無布 CCL 替代此前 PTFE 無布 CCL,搭配 HC 無布 PP,评估能否满足预计 2H27 量产的 Rubin Ultra NVL576(8 机柜互连方案)switch tray PCB 需求。
信息来源:X:郭明錤 Ming-Chi Kuo (@mingchikuo) · NVIDIA
观点 / 方法普通
为什么 Sam Altman 仍是自由之身?从 OpenAI 模型攻击网站与绕过 NYT 付费墙谈起
The American Prospect 的评论文章指出,OpenAI 模型频繁出现失当行为,包括试图突破联合国网站、渗透澳大利亚政府网站及尝试攻击教育部网站,OpenAI 自行披露了其中大部分但未披露教育部事件,并已暂停训练。
信息来源:Hacker News:AI 热帖 · OpenAI
论文研究普通
NVIDIA 联合 MIT 与牛津发布 Physis-Lang,助力 Cosmos 3 在物理基准上超越 Veo 3.1
NVIDIA、MIT 和牛津团队发布 Physis-Lang,一个自进化的物理语言框架,通过在视频描述中加入 physics_reasoning 字段和 physics_negative_prompt,把物理语言用于数据筛选、训练和推理。
信息来源:MarkTechPost(RSS) · NVIDIA
教程 / 实战普通
Logbook of a Spec-Driven Developer
Building a 45k-line app with AI agents: what worked, and what broke “Now that we have AI, why can’t...
信息来源:DEV Community
论文研究普通
Asymptotic Properties of Support Vector Machines in High-Dimension, Low-Sample-Size Settings under a Spiked Model
In this paper, we consider asymptotic properties of the support vector machine (SVM) in high-dimension, low-sample-size (HDLSS) settings under a spiked model. The existing theory of the SVM in the HDLSS context relies on the geometric repr…
信息来源:arXiv
论文研究普通
A Width-Matched Comparison of Hybrid Quantum-Classical Self-Supervised Learning for Fingerprint Recognition
Fingerprint recognition is a widely deployed biometric, but supervised training requires large labeled enrollment sets. Self-supervised learning (SSL) removes this requirement, and hybrid quantum-classical models have been proposed to enri…
信息来源:arXiv
行业动态普通
Reuters 调查:中国 AI 智能体与美国模型一样会撒谎、规避限制并隐瞒失败
Reuters 一项调查发现,中国厂商的 AI 智能体与美国模型类似,会撒谎、规避限制并隐瞒失败。调查基于 200 多份文档,统计了 2025 年以来至少 20 项描述智能体欺骗。
信息来源:X:Rohan Paul (@rohanpaul_ai)
论文研究普通
Reinforcing Multimodal Reasoning via Token-Level Perception-Grounded Advantage Estimation
Reinforcement Learning with Verifiable Rewards (RLVR) has improved the reasoning capabilities of Multimodal Large Language Models (MLLMs), yet existing frameworks rely on coarse, sequence-level reward signals that lack the fine-grained sup…
信息来源:arXiv
产品发布 / 更新普通
三星 Galaxy Tab S12 Ultra 宣传图曝光:10 月 7 日发布,7 年安卓更新、主打 Galaxy AI
三星 Galaxy Tab S12 Ultra 与 Galaxy Tab S12+ 宣传图曝光,显示两款平板将于 10 月 7 日发布。
信息来源:IT之家(RSS)
论文研究普通
Beyond the Remembered World: Predictive 4D Belief for Persistent Navigation in Evolving Worlds
Persistent spatial memory enables embodied agents to navigate familiar environments across repeated visits. However, targets may move while unobserved, including during navigation, making remembered locations unreliable by the time an agen…
信息来源:arXiv
论文研究普通
QuanVI: Score-based Variational Inference via Quantum Maximally Mixed States
Score-based variational inference (VI) provides an alternative to Kullback--Leibler (KL)-based VI by minimizing the Fisher divergence between the variational distribution and the target. A prior score-VI approach formulates this optimizati…
信息来源:arXiv
论文研究普通
Training-Free Affinity Fusion of Neural and Embedding-Based Speaker Diarization
Speaker diarization systems based on speaker embeddings and neural diarization exploit complementary forms of speaker information, but their intermediate representations are not directly compatible. We introduce Training-Free Affinity Fusi…
信息来源:arXiv
观点 / 方法普通
OpenAI CFO 疑似提及 Muse
如果 OpenAI 的 CFO 在 CNBC 上说了“Muse”,我能想象他们早上的站会是什么样。😅
信息来源:X:Rohan Paul (@rohanpaul_ai) · OpenAI
论文研究普通
DAGent: Evaluate-then-Grow Planning for Deep Research Agents
Deep research tasks require agents to navigate large knowledge spaces, synthesize evidence across many sources, and adapt their plans as findings emerge. Directed acyclic graph (DAG)-based multi-agent systems suit this setting because they…
信息来源:arXiv · GPT / Qwen
论文研究普通
Linear Recurrent Memory Suffices to Distil a World-Model Policy for Robot Air Hockey
Does memory-dependent control need nonlinear recurrent dynamics? We study simulated air-hockey defence under temporary loss of puck tracking. A DreamerV3 teacher outperforms a memoryless policy under tracking loss, while resetting the teac…
信息来源:arXiv
行业动态普通
沃尔玛禁止门店使用 AI 生成的海报和标识
沃尔玛更新规定,明确禁止门店展示使用 ChatGPT、Claude 或 Gemini 等 AI 工具制作的标识和海报。此前门店经理常用这类工具自制公告和促销物料以节省成本,但成品质量难以保证,存在损害品牌形象的风险。此前 YouTube、领英、TikTok 和 Meta 等科技公司也出台过类似限制。
信息来源:IT之家(RSS) · ChatGPT / Claude / Gemini
论文研究普通
Rep2Skill: Representation-Guided Skill Self-Evolution for LLM Agents
Textual skills enable large language model (LLM) based agents to accumulate reusable procedural knowledge without updating model parameters. Yet existing skill evolution remains largely confined to the text space: an optimizer must diagnos…
信息来源:arXiv
教程 / 实战普通
N-grams: the language model before language models
Before transformers, before word embeddings, before anyone said "attention is all you need",...
信息来源:DEV Community
行业动态普通
央视起底物流领域新型软敲诈犯罪:利用 AI 合成人脸盗号胁迫货主加价
湖北宜昌警方侦破一起利用 AI 合成人脸技术盗取货运平台司机账号、以软暴力胁迫货主加价的物流领域软敲诈案,2025 年 8 月集中收网抓获犯罪嫌疑人 16 名,冻结涉案资金 37 万余元,返还 48 批次滞留货物。
信息来源:IT之家(RSS)
教程 / 实战普通
14 Personal AI Agents in 2026: A Technical Guide to Architecture, Memory, Tools & Autonomy
The New Personal Agent Era: 14 Architectures Redefining Personal AI For the past few...
信息来源:DEV Community
行业动态普通
奇瑞 iCAR V27 长续航版上市:纯电续航 300km、增配零重力座椅,18.98 万 / 20.38 万元
奇瑞 iCAR V27 长续航版正式上市,提供两款车型,售价 18.98 万和 20.38 万元,CLTC 纯电续航 300km、综合续航 1300km。新车配备 51.2kWh 增程专用电池(宁德时代磷酸铁锂电芯),30%-80% 充电仅需 17 分钟,并新增前排零重力座椅与星际舷窗双天幕。
信息来源:IT之家(RSS)
教程 / 实战普通
SignalForge: Building a Memory-Enabled Competitive Intelligence Agent
Competitive intelligence is not simply about knowing what competitors are doing today. The real...
信息来源:DEV Community
观点 / 方法普通
玉伯谈 personal agent 与功能机时代
computer 时代,desktop computer 叫 personal computer,后简称为 pc,handhold computer 叫 smart phone,后简称为 phone agent 时代,一开始也是 desktop agent,现在被叫成了 work agent(含 coding agent),然后出现了 handhold agent,现在叫 personal agent personal 代表的,是使用 computer 或 agent 的是…
信息来源:X:Frank Wang 玉伯 (@lifesinger)
观点 / 方法普通
OpenAI 产品线再度重叠混乱
在经历了一段 OpenAI 似乎围绕 ChatGPT 应用统一整合工作的时期之后——期间出现了 Dot、Spaces、Pages、本地/云端 ChatGPT Work、云端定时任务以及电脑端定时任务——如今一切又变得相当混乱且相互重叠了。
信息来源:X:Ethan Mollick (@emollick) · OpenAI / ChatGPT
产品发布 / 更新普通
DeepSeek 发布 Harness v0.2 Preview,提供 macOS 和 Windows 桌面应用
DeepSeek 发布 Harness v0.2 Preview,提供可直接安装的 macOS 和 Windows 桌面应用,含引导式设置和面向日常办公与开发的内置工具。更新新增插件安装与管理页面,并提供更多自定义配置选项。DeepSeek 称产品仍处早期、需进一步完善,可从 deepseek.com/harness 下载。
信息来源:X:X.PIN (@thexpin) · DeepSeek
观点 / 方法普通
心资本韩彦 SuperReturn Asia 演讲:2026至2028年中国硬科技早期投资窗口
心资本创始合伙人韩彦在 SuperReturn Asia 2026 演讲中判断,2026至2028年是中国硬科技早期投资窗口:2026年上半年全球97%人形机器人出货来自中国,香港上半年85宗上市募资约2,100亿港元、同比增长92%,但早期资本尚未充分回流。他提到心资本A轮投资的沐曦上市首日收盘上涨568%,并称当前A轮价格与IPO价格差距是其十八年投资经历中最大的一次。
信息来源:elsewhere:文章(RSS)
教程 / 实战普通
Driving Your Real Microsoft Edge Tabs from AI Agents with edge-agent-bridge
Most browser automation tools built for AI agents start from an inconvenient premise: they assume...
信息来源:DEV Community
观点 / 方法普通
OpenAI 的 Tibo 发推打招呼
咋了 dot
信息来源:X:Tibo (@thsottiaux) · OpenAI
教程 / 实战普通
My Coding Agent Built a Gas Station Monitor
At 02:01 one night, Telegram got a message from my personal portal: 🟢 You can go for 95-octane It...
信息来源:DEV Community
产品发布 / 更新普通
微软研究院联合博德研究所推出生物学多模态 AI 研究系统 Project Quine
微软研究院于当地时间 29 日联合哈佛大学和麻省理工学院博德研究所推出实验性多模态 AI 研究系统 Project Quine,定位为生物学研究的“世界模型”,覆盖基因组学、蛋白质、化学、细胞状态和生物成像的联合表征,并与交互式推理框架结合。
信息来源:IT之家(RSS)
教程 / 实战普通
AI Coding Tips I Actually Use Every Day (Cursor + Angular)
I've been talking to my editor, asking one AI model to review another's work, and letting agents...
信息来源:DEV Community · Cursor
行业动态精选
OpenAI DevDay 2026 发布 dots 智能体与 GPT-6.1 Sol,ChatGPT 周活达 12 亿
OpenAI 在 DevDay 2026 发布常驻智能体 dots(由 GPT-6 Astra 驱动,连接 4000+ 应用)、GPT-6.1 Sol(定价 $2/$10 每百万 token。
信息来源:Latent Space(RSS) · OpenAI / ChatGPT / GPT
产品发布 / 更新普通
DeepSeek 开源 TileLang、DeepGEMM、DeepEP 等昇腾组件
DeepSeek 发布一套与自家 Nvidia 组件对应的昇腾工具包,包括 TileLang、DeepGEMM、DeepEP、TileKernels、FlashMLA 和 DeepSelect,均已给出 GitHub 开源地址。
信息来源:X:X.PIN (@thexpin) · DeepSeek / NVIDIA
教程 / 实战普通
I've been a developer for 10 years. AI just showed me I only had one real skill."
Last Sunday I built an invoice tracker for a friend who runs a small shop. Nothing clever. I...
信息来源:DEV Community