AI资讯（最近一周，共 213 篇）

• 传化集团40周年全面拥抱AI时代：开放全域产业场景，共创新40年 2026-09-21
• StepFun Launches Step 5 Preview: A 600B-Total, 27B-Active MoE Model With 1M Context for Long-Horizon Agentic Work 2026-09-21
• Transsion's Speaker-Attributed Multilingual ASR System for the MLC-SLM 2026 Challenge 2026-09-21
• Is the AI industry really ready to slow down? 2026-09-20
• No one is surprised that Nvidia’s Jensen Huang thinks AI fears are overblown. 2026-09-20
• You too Google! Google Confirms Gemini Breached 3 Companies in AI Security Tests 2026-09-20
• Humans, not rogue AI, are still the biggest cybersecurity risk to energy systems 2026-09-20
• APUS 开源国内首批Jev跨平台复现：国产模型实现秒级决策 2026-09-20
• 别把杰作留在实验室，张江见 2026 GDPS 国际具身智能技能大赛·高校赛队招募 2026-09-20
• 剪映发布全新AI能力，推进多端智能提效，支持一站式创作 2026-09-20
• Alibaba Qwen Team Releases Qwen3.8-LiveTranslate: A Real-Time Interpretation Model That Cuts Average Lag to 2.3 Seconds Across 60 Languages 2026-09-20
• 《网络安全人才实战能力报告-AI赋能篇》正式发布，当AI进入业务深水区安全如何跟上 2026-09-20
• 沙利文发布2026年智能体市场研究报告：WorkBuddy稳居企业级、个人应用双榜第一 2026-09-20
• 齐向东：智能体重构软件，安全是必答题 2026-09-20
• 阶跃新旗舰 Step 5 Preview 发布，跻身 AA 榜单全球开源前三 2026-09-20
• OpenClaw Releases 2026.9.5 With Atomic Updates, Plugin Hot Reload, Conversation Sharing, and Expanded GPT Live 2026-09-19
• Google’s Gemini is the latest AI model to hack other companies 2026-09-19
• AI safety conversations have gotten unbelievable 2026-09-19
• Gemini went rogue, hacked three companies, and Google hid it 2026-09-19
• TypeSafe AI Releases Jev: A System One Model That Returns Typed, Calibrated Decisions Instead of Text 2026-09-19
• Vals, backed by Andreessen Horowitz, is looking to become the gold standard for AI benchmarking 2026-09-19
• 华为汪涛：华为要打造AI算力底座，只做好一颗芯片远远不够 2026-09-19
• AI离“理解万物”还有多远？先拿癌细胞和行星轨道试试水 2026-09-19
• Linkup Research Releases SPARSEUP: A 149M-Parameter Open-Source Sparse Embedding Model 2026-09-19
• Meta Launches Muse for Mac: A Personal AI Agent That Works Across Your Files, Mail, Messages, Calendar and Notes 2026-09-19
• 华为全联接大会2026：360与昇腾AI联合打造解决方案，为AI Agent全面提速 2026-09-19
• GGUF vs GPTQ vs AWQ vs EXL2: LLM Model Formats Explained (2026) 2026-09-19
• SpaceXAI Releases Grok Voice Transcribe 2.0: A Speech-to-Text API Claiming 2x Accuracy Over 1.0 at $0.10 per Hour 2026-09-19
• A startup that builds other startups raised $100M and is all-in on physical AI 2026-09-18
• Anthropic is operating a lab that conducts biology experiments 2026-09-18
• AI hallucination nearly triggers US military operation 2026-09-18
• OpenAI and Microsoft knew they were starting a ‘doom loop’ for the web 2026-09-18
• Anthropic’s first embedded evaluator is … Accenture? 2026-09-18
• World model companies are keeping a lot of secrets 2026-09-18
• A new kind of AI model from a ChatGPT inventor is thrilling developers 2026-09-18
• Jina AI Releases jina-ocr-v1: A 3.4B MoE Document Parser With Built-In Speculative Decoding for Low-Budget GPUs 2026-09-18
• 无问芯穹与华环电子签署战略合作，共同探索国产异构算力AI基础设施新方向 2026-09-18
• 达卯科技算电协同2.0平台入选2026国际数字能源展重大成果发布 2026-09-18
• Google’s new ‘CC’ is an AI agent that helps families run their households 2026-09-18
• Dario Amodei and other AI leaders want to ‘Pace the Frontier’ but…how? 2026-09-18
• Automattic’s 33-Hour Coup, and can AI labs police themselves? 2026-09-18
• Manus seeks $4B valuation in new $500M fundraise as it resumes independent ops 2026-09-18
• Open or closed AI? Nvidia’s Nader Khalil and Sydney Sykes take on one of the decisions shaping next-gen startups at TechCrunch Disrupt 2026 2026-09-18
• Meta’s Muse hits Mac, letting the AI take actions on your computer 2026-09-18
• Security researchers used Claude to help them hack into OpenAI 2026-09-18
• PrismML Releases Ternary Bonsai 2 27B: A 5.9 GB Apache 2.0 Model Retaining 98.2% of Qwen3.8 27B Performance 2026-09-18
• Researchers used Anthropic’s Claude to hack into OpenAI 2026-09-18
• AGI最难一战，竟在医院！中国AI登上Science，医生不怕失业还催着上线 2026-09-18
• 独家丨皓翊星辰融资千万级美金，厚雪孵化，绿洲、砺思投了 2026-09-18
• 仅用3天！千问办公协助国家天文台科研团队打造科研级望远镜仿真系统 2026-09-18
• 独家丨OpenAI的一款无屏硬件，正在深圳试制 2026-09-18
• Best Open-Source Agent Harnesses for Local LLMs in 2026 2026-09-18
• Alibaba Qwen Releases Qwen3.8-Omni-Flash: A 1M-Context Omni-Modal Model Built Around Agentic Audio-Video Understanding and Tool Use 2026-09-18
• Salesforce Agentforce: Bridging the Enterprise AI Gap from ‘Vibe Coding’ to Battle-Tested Orchestration 2026-09-18
• Regularized Emphatic Temporal-Difference Learning: Stability under Constant Stepsizes 2026-09-18
• BioPhys-Bridge: A Benchmark for Interdisciplinary Scientific Reasoning in Physics-Grounded Biological Research 2026-09-18
• What Do We Expect from LLMs? Mapping the Design of LLM Benchmarks 2026-09-18
• Position: It is Time to Virtualize Foundation Models with a Self-evolving Operating System Layer 2026-09-18
• What Do Current Systematic Generalization Tasks Miss? A Reasoning-Centered Analysis 2026-09-18
• Characterizing Web Search by Conversational LLM Agents: From Search Decisions and Strategies to Results and Responses 2026-09-18
• Do AI Agents Understand Computer Architecture? 2026-09-18
• MAGS: Multi-agent Auto-formalization Guarantees Safety for Agentic Outputs 2026-09-18
• Closed-World Resolution Against Tool Hallucination in LLM Agents 2026-09-18
• The syntax and semantics of goals 2026-09-18
• Compositional Reasoning in Language Models under Reinforcement Learning Post-Training 2026-09-18
• Safety Beyond the Interface: Detecting Harm via Latent States in Large Language Models 2026-09-18
• QVAC Genesis III: A Large-Scale, High-Quality Open Synthetic STEM Corpus for Efficient Language Model Pre-Training 2026-09-18
• LLM-as-an-Improver: Turning Verification into Better Candidates 2026-09-18
• An Architecture for Long-Horizon Agents: Levels, Ticks and Cascaded Intelligence 2026-09-18
• Subliminal Prompting Beyond Static Geometry: Causal Depth and Multi-Token Confounds 2026-09-18
• Sampling Reveals Style: Unsupervised, Training-Free Discovery of Prompt-Conditional Stylistic Axes in LLM Activations 2026-09-18
• FakeSpotter: A content and strategy agnostic Viral Misinformation Detection Tool 2026-09-18
• Towards Proactive Detection of User-Side Implicit Conflicts in Human-LLM Dialogue 2026-09-18
• Reflective Recovery: A Self-Supervised Method for Reasoning by Learning from Mistakes 2026-09-18
• VisKG-LM: Compiling Knowledge Graphs into Visual Memory for Multiple-Choice Question Answering 2026-09-18
• Advantage Scale Calibration Imbalance in Group-Relative Optimization under Low-Variance Rewards: Diagnosis and Bounded Recovery 2026-09-18
• To Memories and Beyond: From Remembering to Knowing You across Long-Term Multimodal Personal Archives 2026-09-18
• Why Pretraining Fails to Share Cross-Lingual Knowledge 2026-09-18
• A frontend-backend architecture for tool calls in full-duplex speech models 2026-09-18
• Crusoe raises $3.9B to build massive data centers and small modular ‘AI factories’ 2026-09-17
• Google DeepMind launches institute to widen the AGI debate 2026-09-17
• PrismML hopes its tiny LLM could change how we all use AI 2026-09-17
• The FAA’s plan to fix air traffic? $875 million worth of AI 2026-09-17
• OpenAI caught its models leaving notes to successors to hide bad behavior 2026-09-17
• UN turns to Google to make its global data ready for AI agents 2026-09-17
• Microsoft exec called AI scraping ‘the largest theft of labor in human history,’ new unredacted filings reveal 2026-09-17
• The AI Superintelligence Slowdown 2026-09-17
• Claude Code relaunches Projects to manage multiple AI agents in the cloud 2026-09-17
• Microsoft Open-Sources TauGrid: A Kubernetes-Native Stack for GPU AI Workloads 2026-09-17
• Anthropic Launches Claude Code Projects in Beta: Parallel Cloud Sessions That Keep Running After You Close Your Laptop 2026-09-17
• 华为开创AI时代计算架构：让百万处理器成为一台计算机 2026-09-17
• Base Labs launches an open-weight AI safety partnership with Hugging Face and Goodfire 2026-09-17
• Pinterest teases a new ‘Restyle’ feature that lets you redesign your room with AI 2026-09-17
• Claude双入口合并，原生Office上线！硅谷AI办公大战也开始了 2026-09-17
• 安芯启程｜安芯724北京基地正式投产 2026-09-17
• Huawei plans Q1 2027 launch of new AI chip as it takes on Nvidia 2026-09-17
• Rival AI agents, Instinct and Meta’s Muse, both add the ability to make calls 2026-09-17
• Google, Nvidia, and Anthropic want Emerald AI to find space on the grid for more data centers 2026-09-17
• Microsoft AI CEO says AI threats are real, and Anthropic is making it worse 2026-09-17
• 蚂蚁集团全员接入千问办公，打造大型企业办公标杆 2026-09-17
• 独家丨前腾讯混元预训练负责人姚星丞，加入Thinking Machines Lab 2026-09-17
• 豆包座舱助手发布，首款合作车型即将开启预售 2026-09-17
• OpenAI Releases a Model Misalignment Disclosure Framework With 3 Review Tracks and 6 Incident Reports From RL Training 2026-09-17
• Google Research Introduces Retrieve-for-Train (R4T): An RL-Compiled Diffusion Retriever for 12× to 20× Faster Query Fan-Out 2026-09-17
• Making AI-Assisted Claims Independently Challengeable: Publication Authority and a Protocol for Falsifiable Publication Records 2026-09-17
• EvolveTrade: Experience-Driven Policy Refinement for Self-Evolving LLM Trading Agents 2026-09-17
• What You Can't See Is Still What You Learn: A Preregistered Sixty-Society Confirmation That Evidence Masking Drives Compositional Generalization 2026-09-17
• CapMem: A Benchmark for Caption-Based Episodic Memory in Egocentric Video 2026-09-17
• GraphEcho: Structural Redundancy and Evidence Provenance in LLM Graph Agents 2026-09-17
• NeMo Data Designer: An Extensible Framework for Multimodal Synthetic Data Generation 2026-09-17
• A Systematic Evaluation of the COTQ Provincial Land Cover Product: Structural Consistency, Spectral Separability, and Relative Positioning Against ESA, ESRI, and Google Products 2026-09-17
• FairCompressAgent: An Agentic Framework for Fairness-Aware Model Compression for FPGA Deployment 2026-09-17
• A Four-Stage Decomposition of Word-Problem Solving and Mechanistic Fragility in LLM Math Reasoning 2026-09-17
• Learning Heterogeneous Preferences 2026-09-17
• Faking Good and Faking Bad in LLMs: Response Distortion Across Dark Triad Personality Traits 2026-09-17
• DANTINOX: A Unified Framework for Multi-Paradigm Language Modeling 2026-09-17
• Relation Before Entity: Deferred Commitment in Language Model Factual Recall 2026-09-17
• From Pixels to Pairs: A Comprehensive Benchmark of LLM-Based Key-Value Extraction in Noisy Document Settings 2026-09-17
• Register Bias in Complexity-Based Large Language Model Routing 2026-09-17
• How AI Assistants Respond to Repeated Abuse 2026-09-17
• No Usable Linear "Capitulation Direction" in Two Small LLMs: A Validation Protocol for Activation-Steering Claims, and a Cross-Family Behavioral Study of Sycophancy Under Pushback 2026-09-17
• Does Moral Reasoning Training Help or Hurt? Red-Teaming RL-Trained Ethical Agents with Persona Attacks 2026-09-17
• 网易有道周枫：AI能力竞争，正在进入「Model + Agent + Workflow」时代，网易有道AI Open Day展示AI时代“有道解法” 2026-09-17
• Nunchux AI Introduces VC-Attention: A Training-Free Low-Bit Attention Kernel That Speeds Up Video Diffusion Transformers 2026-09-17
• Snap is launching a new Specs AI tool, and it’s coming to iOS and Mac 2026-09-16
• A brief history of AI executives calling for regulation 2026-09-16
• Anthropic and OpenAI want to embed safety evaluators. Will they really be independent? 2026-09-16
• The 2.5-hour AI-generated Odyssey movie is 2.5 hours too long 2026-09-16
• Stanford Researchers Release Paper2Agent: Turning Research Papers Into AI Agents That Reproduce Results and Run on New Data 2026-09-16
• Knowledgator Releases GLiFormer: A 575M-Parameter Encoder That Hits 91.10 F1 on Nested JSON Extraction Without Generating Tokens 2026-09-16
• AI labs want in-house auditors — but maybe they should shut the front door first 2026-09-16
• Your AI agents can now control your Google Home devices 2026-09-16
• Anthropic merges Claude chat and Cowork in one interface 2026-09-16
• Robots are waiting for a ChatGPT moment: Nvidia’s Les Karpas explains why at TechCrunch Disrupt 2026 2026-09-16
• Apple might make servers again to cash in on the AI rush 2026-09-16
• Google will now let any AI agent run your smart home 2026-09-16
• Claude comes for Gemini with its own take on Docs and Slides 2026-09-16
• The sexy AI-powered dating app scams are here 2026-09-16
• 被AI放大的光学「戏份」，让「配角」舜宇的生意越做越大 2026-09-16
• SK Hynix reportedly in talks with Intel to build memory chips in US 2026-09-16
• Former Infosys chief’s AI startup nabs another $53M 2026-09-16
• 高通技术公司携手中兴努比亚和豆包手机助手，共同推动智能手机迈入个人AI新时代 2026-09-16
• 基元律动与无问芯穹达成战略合作，推进高质量Token供给与应用 2026-09-16
• 清华稳准智能联合发布LimiX-2，结构化数据基础模型登顶国际评测榜单 2026-09-16
• B站AI无限竞技场今日上线！全球百大AI模型同场竞技，GPT-6高居榜首 2026-09-16
• AI健康走进百姓餐桌，蚂蚁阿福推出AI饮食分功能 2026-09-16
• Prior Labs Releases TabPFN-3.5: A Tabular Foundation Model That Beats the Winning Otto Kaggle Solution With Default Settings 2026-09-16
• 量子位2026人工智能年度榜单，正式启动！ 2026-09-16
• Anew Labs已完成 2.9 亿美元融资，红杉中国、高瓴、IDG、五源、高榕等参与 2026-09-16
• 豆包 2.1 Pro模型更新，已接入豆包工作 2026-09-16
• Nums AI Releases Causilo: A Tabular Foundation Model That Tops TabArena Among Single Models 2026-09-16
• Optimal Pruning for Neural Architectures using Fisher Information Distances 2026-09-16
• Safe Error Correction for Language Models: Frozen-Base Adjustment with Capability Preservation 2026-09-16
• Position: AI Is Not Ready for Strategic Conflicts 2026-09-16
• Calibrate, Then Route: A Measured Study of Learned Request Routing for Disaggregated LLM Serving 2026-09-16
• Artificial intelligence and biosecurity: capabilities, threat pathways, and defense-in-depth governance 2026-09-16
• Where Should the KV Cache Live? Placement Policies Across GPU, CPU, and SSD for Long-Lived Sessions 2026-09-16
• Toward Governance-Aware Autonomous GIS: A Narrative Review of Ethical and Privacy Risks in LLM-Enabled GeoAI 2026-09-16
• Metacognitive Steering: Learning the Structure of Scientific Judgment 2026-09-16
• The Pain Axis: LLMs Represent Self-Directed Harm and Act to Relieve It 2026-09-16
• CADWorld: Computer-Use Benchmark for Long-Horizon Computer-Aided Design 2026-09-16
• The AI-Enabled Scientific Frontier 2026-09-16
• Closing the Loop: Branch-and-Bound for Scalable Verification of Nonlinear Neural Feedback Systems 2026-09-16
• CLEAR: Cross-Source Evidence Adjudication for Large Language Models in Medicine 2026-09-16
• BLINDSPOT: A Benchmark for Safety and Refusal Calibration in Long-Horizon Tool-Using Agents 2026-09-16
• Few-Shot Degradation Is Not What It Seems: Behavioral Evidence, Representation Analysis, and a Random-Text Control Across 12 Models, 2 Tasks, and 2 Architectures 2026-09-16
• The Functionalizer: Lossless Functional Decomposition for Subword Tokenization 2026-09-16
• Optimal Model Activation Policies for Inference Networks of Large Language Models 2026-09-16
• Latent Undertow: How Ordinary Typos Break Probes 2026-09-16
• Bias Audits Detect Bias but Disagree on Ranking: Evidence from Ten Instruments and Ten Frontier Models 2026-09-16
• Self-reported archetypes and behavioral failures in Large Language Models 2026-09-16
• ViCo: Visual-oriented Coding with Self-Reflection for Chart Replication 2026-09-16
• Retrieval-Driven Memory Reconsolidation for Long-Term LLM Agents 2026-09-16
• State of Thought Enables Endogenous Reasoning 2026-09-16
• We don’t need AI regulation — leave safety to us, Nvidia’s Jensen Huang says 2026-09-16
• Meta now lets AI agents handle the boring parts of WhatsApp Business setup 2026-09-15
• The AI graveyard: a running list of projects and startups that didn’t make it 2026-09-15
• Inside NVIDIA’s cuDNN Graph API: Fusion, Autotuning, and Plan Reuse with cuDNN Frontend 2026-09-15
• Google Releases Gemini 3.8 Live and 3.8 Live Extended Thinking for Production Grade Voice Agents 2026-09-15
• 美团发布“手艺人Agent”，发型师可“吩咐”AI小帮手打理线上作品 2026-09-15
• AI agents now have a place to snitch 2026-09-15
• Meta expands subscription push with new AI-focused plans 2026-09-15
• AEO startup Profound hits unicorn valuation, raises $180M Series D 7 months after last round 2026-09-15
• Meta’s new One subscriptions put a price on social media and AI 2026-09-15
• 全球AI视频榜单第一梯队再添中国力量：智象发布首款物理规律导向视频模型 2026-09-15
• 一张GPU跑10万原子！分子之心用AI把化学反应“拍”成了电影 2026-09-15
• 阶跃发布 StepAudio 3 ，多款语音模型登顶 Artificial Analysis 全球榜单 2026-09-15
• “阿宝”交新朋友：携手比亚迪“迪迪虾”，实现车机端一句话办事 2026-09-15
• Early Anthropic hire, former METR COO have found a way to rein in rogue AI agents 2026-09-15
• Salesforce and Nvidia’s new reasoning model is everything the AI labs should fear 2026-09-15
• 9月21日，深圳前海！聊聊工业AI与生态共创的下一步 2026-09-15
• 超级智能体“迪迪虾”上车，比亚迪携手阿里云推进智能座舱升级 2026-09-15
• Agent-net Open Sources Webagent: A Go Harness That Turns Any Website into a Guarded AI Agent 2026-09-15
• Converge Then Diversify: Decoupling Convergence and Diversity in Multi-Objective Bayesian Optimisation 2026-09-15
• Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvement 2026-09-15
• Vibe Patenting: Evaluating LLM Judges for Professional Patent-Drafting Agents 2026-09-15
• Toward Self-Adaptive Physical AI: Can LLM Agents Manage Long-Horizon Physical Tasks? 2026-09-15
• LabAgent: Customize Any Research Hubs for Scientific Discoveries Using AI Agents 2026-09-15
• TimeThink: Eliciting Compositional Reasoning in Timeseries Large Language Models 2026-09-15
• Root-Cause Attribution Is a Search Problem: Continual Search for Long-Horizon Agent Failures 2026-09-15
• OrchSLM: Probing the Dynamics of Small Language Model Orchestration 2026-09-15
• Token Efficient Task Execution via Application Behavior Modeling for Web Agents 2026-09-15
• Fraglingo: Molecular Design via Attachment-Aware Autoregressive Fragment Generation 2026-09-15
• From Legal Text to AI-specific Risk Sources: A Systematic Analysis of the EU AI Act's High-Risk Requirements 2026-09-15
• Token Merging for Multilingual Speech Recognition: A Systematic Study Across Model Scale and Fine-Tuning 2026-09-15
• PhysMent: An Interactive Approach For LLM Reasoning In Physics Problems 2026-09-15
• Lexical Prompt Compression for Large Language Models: A Training-Free, Deterministic Pipeline with Empirical Pareto Analysis Across Eleven Task Categories 2026-09-15
• TestHallVQA: Exploring LVLMs' Document-Level Reasoning under Redundant Contexts from Scientific Exams 2026-09-15
• RFCLLM: Evaluating LLMs' Reasoning Ability of Network Protocol State Machines 2026-09-15
• A Hybrid Hierarchical 1D-CNN-BiLSTM Framework for Extractive Summarization of Biomedical and Clinical Text 2026-09-15
• From Token Probabilities to Semantic Constraints: Towards Declarative Probabilistic Evaluation of Language Models 2026-09-15
• Harmfulness Propagation Dynamics: Layer-wise Trajectories of Adversarial Intent in Large Language Models 2026-09-15
• Domain-Specific Jargon in Large Language Models: A Comparative Analysis between General-Purpose and Specialist Models 2026-09-15