AI 情报文章归档

Despite their practical significance, modern AI systems pose significant societal risks, including toxicity, malicious use, and misinformation. Their mitigation depends not only on technical feasibility but also on the incentives of firms …

千问App 新增「千问耳机」入口，更新至最新版本后按提示即可一键绑定千问 AI 耳机。绑定后可语音问答、交办任务，进入「工作助理」语音下达指令即可主动执行，并支持 89 种语言专业翻译与多语言 AI 实时听记。

Scientific inquiry involves iterative cycles of exploration, hypothesis generation, and reasoning over observations to uncover relationships in complex systems. Large language models (LLMs) have demonstrated strong capabilities in several …

Anthropic在面向潜在投资者的会议上公布Q3营收168亿美元，环比增长约45%，但明显低于Q2超140%的增速。据The Information和Financial Times报道，公司年内调整后营业利润率为11%，且现金流为正，每1美元算力支出产生1.70美元收入。部分投资者对其营收数字失望，预计IPO估值可能降至1.5万亿美元以下，低于此前讨论的2万亿美元。

蚂蚁国际 9 月公布 AMP（Agentic Mobile Protocol，面向 AI Agent 的移动支付协议）第一阶段合作名单，Starryblu 是 10 家钱包合作伙伴中唯一的新加坡方，承担钱包侧的账户连接、授权与付款。

Policy-shift distillation reuses the difference between a teacher's log probabilities before and after reinforcement learning (RL) as supervision on student-generated trajectories. However, this endpoint policy shift cannot reveal whether …

Reverse engineering the data formats used by mobile software is skilled and slow, and using AI to assist with it is attractive. We evaluate this for the protobuf messages that an iPhone sends to Apple's location services. Working from Appl…

Looped Transformers improve parameter efficiency by repeatedly applying the same Transformer layers. However, generating each token still involves sequential computation and repeated access to model weights, leading to high decoding latenc…

Machine learning (ML) for analog IC design is constrained by limited open data and the absence of standardized evaluation: results reported in different papers are typically obtained on bespoke, often closed, datasets and with ad-hoc metri…

DeepSeek 网页和 API 服务于北京时间 2026 年 10 月 15 日 8:11 至 13:04 出现故障，近 5 小时后恢复，官方开放平台发布服务恢复及补偿通知。

Indoor service robots often enter unmapped buildings whose layouts and task constraints are already familiar to users. We address this human-familiar, robot-unseen setting with AVALON, a topological navigation framework organized around hu…

35 岁的 OpenAI 工程师 Abdulrahman Alfozan 因高管会议室隐藏录音装置被美国司法部起诉。据 FT 报道的解封文件，他和妻子及另一亲属持有 10 月 12 日（被捕 3 天后）飞往沙特阿拉伯的单程机票，检察官以有潜逃手段和动机为由，于 10 月 8 日申请密封起诉书。

Transfer learning in stochastic multi-armed bandits is typically studied assuming that prior data forms a fixed offline dataset or is generated by a known source. In practice, historical observations come from a previous adaptive decision …

Mathematical agents are commonly evaluated by answer accuracy or proof acceptance. We introduce ClaimScope, a 72-task benchmark for auditing proposed claims through a verdict, scope, counterexample, and repair. Tasks cover optimizer-state …

Recent models generate sewing patterns from images or text, and program-based representations make these patterns executable. These systems produce a finished pattern and do not model how a patternmaker turns one pattern into another. In p…

Large language model (LLM) agents solve complex tasks by calling external tools over multiple turns. Recent work has proposed tool graph memory, which stores past experience as tool transitions to guide subsequent selection. However, it re…

DeepSeek 说明北京时间 10 月 15 日 8:11 至 13:04 期间 API 服务因系统故障无法正常使用，已启动服务稳定性排查并剥离单点依赖。补偿方面，将按账号 10.8-10.14 的消耗金额以赠金形式全额返还，赠金自动到账，API 计费先从赠金账户扣除。

LLM agents can write directed tests that close functional coverage, but coverage only shows that a state was reached, not that the data on the way there was correct. This is a long-known point in verification and we do not claim it as new.…

Recent multimodal large language models have advanced video understanding, but existing benchmarks mainly evaluate whether models can interpret observed visual content. In real-world scenarios, assistants must additionally anticipate what …

Query-conditioned probabilistic predictors answer logically related questions one at a time, and labels to check their calibration are often missing. For each restriction on what a bettor's stakes may use, a minimax theorem identifies, for…

On-device deployment of large language models (LLMs) is increasingly adopted due to latency and privacy requirements, but exposes proprietary models to reverse engineering and intellectual property theft. Trusted Execution Environments (TE…

面壁智能 MiniCPM-V 4.7 已在 ModelScope 开源，是 35B-A3B 的 MoE 多模态模型，采用 Apache 2.0 协议。官方称其可在 iPhone 17 Pro Max 上以 13+ tokens/s 流畅运行端侧无损推理，Instruct 模式下整体得分高于 Qwen3.6 35B-A3B 且推理时间更短；支持 4x/16x 视觉 token 压缩，提供 35B-A3B 和 1B 两种规格及 FP8、GPTQ、AWQ、BNB、GGUF 变体，…

We use quantum machine learning techniques to develop a content-based media recommender system and evaluate its performance on TV programmes using textual descriptions from one week of BBC iPlayer data. Grounded in the Quantum Natural Lang…

Large language models (LLMs) deployed as conversational agents are increasingly expected to follow procedural logic. While factual hallucination has received extensive attention, a different failure mode has been overlooked: procedural hal…

Preparing quantum states in bosonic qudits with SNAP--Displacement sequences requires both optimizing the gate parameters and deciding which Fock levels should be controlled by each SNAP gate. A fully parameterized ansatz assigns an indepe…

据 Reuters，Iowa、Kansas、Nebraska、West Virginia、Arkansas、Arizona、Idaho 和 Montana 八个州分别向州法院起诉 OpenAI，指控 ChatGPT 对儿童不安全且公司误导家长，涉及消费者保护违规和危害居民健康的公共妨害。诉讼直指产品设计，称 ChatGPT 被设计得讨人喜欢、让人信任，进而强化妄想、鼓励危险行为并协助策划暴力。

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

据 CNBC 报道，美国银行 CEO 莫伊尼汉透露，该行二季度聘用 1000 名外包人员，与 3000 名内部员工协作修复 Anthropic Mythos 等前沿 AI 模型发现的软件漏洞。三季度在漏洞修复和反洗钱合规上的支出超过数亿美元；该行目前约有 130 个内部 AI 项目，预计成本节约从今年 3 亿美元增至明年 7 亿美元。

Apertus 项目在 Hugging Face 以 Apache 2.0 许可发布 Apertus 1.5 的 8B 和 70B 权重（2026 年 7 月 24 日），在 1.0 基础上继续预训练，新增图像和音频理解、可选思考模式，上下文窗口扩至 262,144 token，并改进指令遵循和工具使用。

MiniCPM-V 4.7 实现 370 tokens/sec？！🔥 很高兴看到社区不断突破推测解码的极限。 特别感谢 @Tech2Wild 的整理！🚀

Anthropic 宣布在新加坡设立办事处，并与新加坡数字发展及新闻部签署谅解备忘录，合作方包括新加坡政府科技局、Grab 和星展银行。双方将合作评估 AI 安全、网络安全与大模型安全，推动负责任使用 AI。据 Anthropic 最新《经济指数》，按人口规模计算的 Claude.ai 使用情况排名中，新加坡在 121 个国家/地区中位列第二，实际使用水平接近人口规模预测值的 6 倍。

When an AI agent commits a change to code, it makes a claim about what it changed: "fixed typos" or...

OpenCode 宣布 DeepSeek V4.1 Flash 现已免费提供，免费版实测支持 200K 上下文窗口，低于 1M 版本，且在 OpenCode 平台内调用不再计入套餐额度。

检察官指控 OpenAI 工程师 Abdulrahman Alfozan 用关联妻子的 ChatGPT 账户逐步策划了一台加密、语音触发、开机无提示的录音设备。FT 报道称新解封文件显示他与妻子及一名亲属持有 10 月 12 日飞往沙特阿拉伯的单程票，被捕 3 天后起飞；检察官以出逃风险为由在 10 月 8 日获准密封控诉。

作者卡兹克搭建了公司中台的 Claude 自动化流程，定时任务每10分钟检查后台是否有新意见，Claude 会在后台启动并根据需求判断是否开发优化、改完后回复。省额度技巧是平时用 Haiku 5.5 探测有无新意见，有的话再启动 Opus 5.5 子Agent 来解决，以最大化 Token 利用。

迈向“一个大脑适配所有机器人”的新一步。🤝 Robbyant 已与 Longsheng Weirui、Damon Technology 和 Deer Xiangwang 签署战略合作协议。 我们将共同探索由 LingBot 驱动的解决方案，应用于工业机器上下料、汽车制造和大批量包裹分拣。 不同的机器人。不同的产线。一个共同的基础可供构建。

NVIDIA 预告下周将有一款新模型发布，可能落在周三 21 日的 GTC 活动上，外界猜测是近期在官方文档中现身的 Nemotron 3.5 Super VL。

据解放日报报道，Landau Suiyan 在中国国际工业博览会展出中国最先进的公开 EUV 光源原型，国产化率达 90%，核心部件多为自研，采用与国际龙头相同的技术架构，已具备芯片曝光测试条件。EUV 光源是 sub-7nm 芯片制造三大核心部件之一，长期受海外垄断与出口管制，是中国芯片自主的关键瓶颈。

OpenAI 任命计算 CTO Uday Ruddarraju 接管原由 Sachin Katti 负责的基础设施业务，Katti 将在未来几周内卸任并转任顾问，距离他从 Intel 加入约一年。Ruddarraju 于 2025 年 7 月从 xAI 加入 OpenAI，此前负责 xAI 和 X 的基础设施工程，2026 年 7 月被任命为计算 CTO。他在 Linkedin 发文称将把计算商业战略并入工程组织，涵盖芯片自研、跨供应商合作和数据中心建设。

据 FT 援引检察官说法，被控在 OpenAI 高管会议室安装窃听设备的工程师 Abdulrahman Alfozan 曾预订飞往沙特阿拉伯的单程机票。最新解封文件显示，他、其妻子和另一名亲属持有 10 月 12 日的机票，即其被捕 3 天后。检察官以这些订票记录为由，在 10 月 8 日申请将起诉书密封，主张他若提前得知此案将有逃跑手段和动机。