2026-09-17 科技日报

扫描 552 篇候选内容 · 覆盖 650 个信息源 · 纳入 535 篇

📌 今日要点

• AI labs want in-house auditors — but maybe they should shut the front door first
• Your AI agents can now control your Google Home devices
• Anthropic merges Claude chat and Cowork in one interface
• Robots are waiting for a ChatGPT moment: Nvidia’s Les Karpas explains why at TechCrunch Disrupt 2026
• Threads’ new features let podcasters promote shows and reach listeners
• 刚刚，全球首个开源 Agent 会话平台发布！老板、同事和我共享一个 AI

📝 科技简讯

• AI labs want in-house auditors — but maybe they should shut the front door first — TechCrunch AI
• Your AI agents can now control your Google Home devices — TechCrunch AI
• Anthropic merges Claude chat and Cowork in one interface — TechCrunch AI
• Robots are waiting for a ChatGPT moment: Nvidia’s Les Karpas explains why at TechCrunch Disrupt 2026 — TechCrunch AI
• Threads’ new features let podcasters promote shows and reach listeners — TechCrunch AI
• 刚刚，全球首个开源 Agent 会话平台发布！老板、同事和我共享一个 AI — 新智元
• Claude 王座失守！OpenAI 一夜反超，神秘 GPT‑6 Sol 偷跑 — 新智元
• 自进化社媒 Agent！浙大 x 北大：从多模态任务编排到反馈驱动 — 新智元
• 刚刚，中国队 9B「全能王」开源登场！横扫 8 项全球通用第一 — 新智元
• 刚刚，Gemini 3.8 Live 上线！ — 新智元
• 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 — ArXiv CL (cs.CL)
• The Functionalizer: Lossless Functional Decomposition for Subword Tokenization — ArXiv CL (cs.CL)
• Optimal Model Activation Policies for Inference Networks of Large Language Models — ArXiv CL (cs.CL)
• Single Document Extractive Summarization using Domination in Hypergraph — ArXiv CL (cs.CL)
• Latent Undertow: How Ordinary Typos Break Probes — ArXiv CL (cs.CL)
• How to Make Linear Regression Survive Outliers — Towards Data Science
• Silent Broadcasting Can Ruin Your Model — Towards Data Science
• The KV Cache Tax: Why Inference Servers Run Out of Memory Before Compute — Towards Data Science
• The N Squared Pizza Problem — Towards Data Science
• MechReason: Benchmarking Multi-Image Multi-Hop Reasoning in Mechanical Engineering — ArXiv CV (cs.CV)
• DenseFace: Bias Mitigation in Face Recognition via Density-Aware Probabilistic Matching — ArXiv CV (cs.CV)
• Hyperbolic Contrastive Learning with Entailment for Spatial Transcriptomics — ArXiv CV (cs.CV)
• SceneBench: A Hierarchical Benchmark for Vision-Language Understanding of 3D Scenes — ArXiv CV (cs.CV)
• ProtoLIP: From Sentence-Level to Object-Level Evidence Disentanglement — ArXiv CV (cs.CV)
• EU president warns AI agents “escaping their environment” are just a preview of what’s coming — The Decoder
• Apple is reportedly building an enterprise AI server with its own M8 Ultra chips — The Decoder
• Google Deepmind launches interdisciplinary institute to tackle the big questions around AGI — The Decoder
• Anthropic merges Claude Chat, Cowork, and more into a single product — The Decoder
• Former OpenAI researcher builds an AI model that judges options instead of writing text — The Decoder
• Mozilla’s new Smart Window assistant runs on Mistral’s models — The Decoder
• Political opposites unite in Washington to rein in AI — The Decoder
• Bi-MoDe: Bilateral Control-based Imitation Learning via Modifier-Conditioned Decoding for Modulation of Execution Speed and Contact Intensity — ArXiv RO (cs.RO)
• MR-GLi: Mixed Reality-Based Gripper-Linked Overlays for Underwater Robot Arm Teleoperation via Bilateral Control — ArXiv RO (cs.RO)
• World-Action Models for Robot Learning and Control: A Survey — ArXiv RO (cs.RO)
• AssemblyGrid v1: A Benchmark for Multi-Robot Production with Temporary Coalitions, Local Information, and Geometric Constraints — ArXiv RO (cs.RO)
• Structure-Preserving Quantum Circuit Architectures for Robot Kinematics — ArXiv RO (cs.RO)
• 通用能力不打折，空间具身智能断层领先！ZDTaichu5.0-9B 国产开源，跻身全球多模态第一梯队 — 量子位
• AI 大模型工场 2026 AI 产业生态大会今日举办，大咖同台共探智能生长与产业共生 — 量子位
• 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 — 量子位
• 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 — 量子位

📊 数据概览