AI 情报文章归档

Text-to-image models have made remarkable progress in visual fidelity and instruction following, yet complex compositional generation remains challenging. Existing agentic approaches address this limitation through iterative generation and…

蚂蚁 inclusionAI 在 GitHub 发布 BiDex-100，一个覆盖 106 个仿真任务的双臂灵巧操作基准，用于评测视觉-语言-动作（VLA）策略。该基准包含 8,080 条成功验证演示、6.25M 帧、58 维统一动作空间，并提供匹配的 ID/OOD 评测协议与 LeRobot v2.1/v3.0 数据集。

Participant-disjoint cross-validation prevents direct participant overlap, but it may still produce optimistic estimates when acquisition site and diagnosis are structurally dependent. We examined Parkinson's disease classification using 1…

三星电子正在开发 zNAND-T 与 zNAND-I 两款 AI 负载优化 NAND 闪存，加上此前已公布的 zNAND-O 共三种型号。zNAND-T 的 "T" 代表吞吐量，类似闪迪和 SK 海力士的 HBF 概念，采用 TSV 堆叠 NAND Die 优化读写带宽；zNAND-I 的 "I" 指 IOPS，强调少量数据频繁读写场景，如 KV Cache 卸载。

Time series (TS) anomaly detection (AD) lacks a universal detector, with the best choice determined by the series rather than by the algorithm. Existing methods therefore either select one detector for each series or average a pool of them…

The daemon crashed, came back up, and recorded nothing. No gap, no warning, no log. The code was...

This is a submission for the Kaggle Benchmarking Challenge. Team: @g00ds0ul · @joyy-dev Benchmark:...

Embedded platforms increasingly host safety-related streaming functions next to less critical ones. Whether a safety architecture pattern, such as a doer/checker pair, protects a function depends on the cores that run its components. Their…

哇 https://x.com/i/grok/share/bGVnYWN5_22468d8a-a1a4-4391-9796-cda7cf1937b5

Spatial transcriptomics jointly captures molecular state, tissue morphology, and spatial organization, yet spatial-domain methods are typically compared only at late training checkpoints, obscuring differences in convergence efficiency. We…

黑五的账算不明白？ 输入 5,000 行，输出 3,000 笔真实订单。 我们的财务与会计专家揪出了 636 万美元的重复项和退款 👇

Language models trained only on text from before a given year are useful for studying hindsight and historical language, but the open collections of pre-modern Chinese carry no publication dates and unreliable dynasty labels. We present Lü…

Freely swiveling casters are widely adopted in industrial AMRs/SDVs for their mechanical simplicity and load capacity. However, caster misalignment during low-speed or high-curvature maneuvers induces strong contact nonlinearities, widenin…

TikTok CEO 周受资在新加坡星展集团首届亚洲未来前瞻论坛上表示，市场对 AI 短期发展的能力可能有所高估，却低估其长期影响，未来两至三年 AI 有望推动物理学、生物科技等领域出现重大突破。他主张采取“人类参与其中（Human-in-the-loop）”模式，在高风险工作中不应完全仰赖自动化，并鼓励各行业亲自使用最先进的 AI 模型以了解其优势与限制。

Indigo 在 Indigo Talk EP52 中提出，AI 把写软件的成本压到接近零，软件将像牛奶一样成为日用品，价格贴着成本走。软件从千人千面走向为个人目标定制、用完即弃，兼容性不再重要。@turingou 表示重心已从造软件转向想清楚软件要达成什么目的，未来值钱的是目的本身，能绕开软件直达目的就不需要软件。

A model may select the correct answer yet fail to construct it through a sequence of local decisions. We introduce Choice2Seq, a diagnostic benchmark that measures correctness retention across complete-answer selection and sequential const…

The theme for the first Hacktoberfest DEV Challenge is out, and it is Build for a Friend: pick one...

作者发布斯坦福课程 CS329A「Self-Improving AI Agents」第一讲 Course Overview 的学习笔记，课程 9 讲视频已发布在 YouTube，第一讲获 52 万观看。

Existing cumulative regret guarantees for Bayesian optimization (BO) do not necessarily characterize the quality of the algorithm's final output. This gap is critical under noisy evaluations, where the goal is to return a single high-quali…

百度世界 2026 将于 11 月 12 日在北京·国家会议中心二期举办，大会主题为“下一幕，智能组织”，官网已开启售票。官方预告本届大会将带来重磅嘉宾与企业级 AI 产品集中亮相，并设 12 场分论坛、30+ 公开课及 6000m² 展区。

有意思。 物理学家 Alexander Unzicker 博士称数学家们针对 OpenAI 700 篇数学预印本的声明是“荒谬至极”，并将大部分批评指向 Terence Tao，认为是他转发了该声明。 ---- （完整视频见“Unzicker's Real Physics”YouTube 频道，链接在评论中）

周 末 说明：该推文正文仅为“W E E K E N D”（周末），不含任何 AI/ML/LLM 相关信息、主体或动作，无法按规则生成 10-15 字且点明核心主体的中文标题，也无法产出有意义的翻译摘要。

Flagship large language models (LLMs) offer rich general knowledge and strong reasoning capabilities, but their high inference cost and latency limit their use for large-scale data labeling. We study zero-shot text classification with sema…

I sit somewhere in the middle of the AI hype cycle. Not a hater nor a fanatic. I’ve been trying to...

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass What I...

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass I keep...

试试 Grok @Bot

Real-world studies (RWS) analyze routinely collected healthcare data, such as electronic health records and insurance claims, to estimate treatment effects, monitor safety, and identify risk factors outside randomized trials. Automating RW…

Recent multimodal large language models (MLLMs) achieve strong results on video question answering, yet often fail to recover the temporal structure that makes a video coherent. They may merge neighboring events, miss single-event progress…

This manuscript presents DeepThink, a knowledge-addressable framework for recurrent language modeling. A Knowledge Bank stores typed Knowledge Units, a Knowledge Activator selects units from the current execution state, Knowledge Assembly …

We propose the task of Interactive 4D Volumetric Liquid Forecasting: predicting the spatiotemporal evolution of a future dense 3D liquid velocity field conditioned on how a solid object moves through it. This task provides an upstream phys…

Large language models (LLMs) have demonstrated remarkable capabilities in solving complex tasks in coding, mathematical logic, and iterative reasoning, especially by coordinating effectively within multi-agent systems (MAS). However, langu…

A fruitful approach towards understanding generalization in machine learning involves characterizing how models separate generalizable structure from idiosyncratic or even corrupted features of training data. Fisher information has emerged…

阶跃 Step 5 Preview 模型上线腾讯 WorkBuddy，2026 年 10 月 12 日至 10 月 25 日在 WorkBuddy 独家限免。该模型采用稀疏 MoE 架构，总参数量 600B、激活参数 27B，支持 100 万 Token 上下文窗口，原生支持文本与视觉输入，面向 AI 编程、软件工程、专业知识工作、金融等场景，并将于 10 月 15 日开源完整权重。

Interactive-agent evaluations commonly rely on standardized LLM-based user simulators that exhibit broadly cooperative behavior, potentially overlooking failures that arise under diverse but realistic user interactions. This can lead to ov…

数字生命卡兹克撰文分析AI驱动逆向工程热潮：开发者Brandon Thomas因不满Adobe订阅规则，用Claude Opus 5.5以Rust逆向重写了Photoshop、Illustrator等7款软件的开源替代品PhotoCraft、VectorCraft等，不到10天单PS项目获4.6万Star。

令人沮丧的是，这个平台上普通的回复机器人已经把它的回复垃圾角度转向了提出有趣的问题。 这感觉像是真正的智识互动，这让它比仅仅用"这才是人们忽略的部分"来改写帖子更糟糕。

阶跃 Step 5 Preview 已在腾讯全场景 AI 办公工作台 WorkBuddy 上线，用户可通过自然语言指令让其自主拆解任务、调用工具。该模型将更强的推理与 Agent 能力带入深度研究、信息检索、数据分析、文档与 PPT 生成及多步骤复杂任务执行等办公场景。2026 年 10 月 12 日至 10 月 25 日，Step 5 Preview 在 WorkBuddy 开启独家限免。

Why do LLMs confidently invent APIs that have never existed in the history of the universe? It is not a memory defect. Agents hallucinated because they were committed to the prompt—and Fernando Flores solved the root cause in 1986.

宇树科技创始人王兴兴在第二十六届中国国际工业博览会上表示，过去十几年他犯的最大错误就是对 AI 还不够相信，如今对 AI 的信仰与日俱增。他透露 GPT-6 发布对具身智能行业认知影响很大，团队已将其部署至机器人测试且效果不错，认为行业临界点可能一两年内到来。