Tech News

Google DeepMind

Các nhà khoa học tạo ra một bản đồ số chi tiết giúp máy tính nhận diện và phân loại từng góc nhỏ rừng cây trên Trái Đất qua hình ảnh vệ tinh.

Hugging Face

Mô hình trí tuệ nhân tạo mới này có thể nhìn vào các bảng dữ liệu phức tạp của doanh nghiệp và tự động đoán trước sự việc mà không cần phải dạy nó từ đầu.

Hugging Face

Công cụ này giúp máy tính hiểu và nhớ được hình ảnh chụp từ vệ tinh của cả Trái Đất, giống như thu nhỏ toàn bộ hành tinh vào một cuốn bản đồ thông minh có thể tìm kiếm mọi thứ.

Hugging Face

Việc tối ưu hóa cách một chương trình AI cắt nhỏ chữ cái giúp tiết kiệm rất nhiều chi phí chạy máy tính mà không làm giảm độ thông minh của nó.

Google DeepMind

Công cụ này giúp các thiết bị nhỏ như điện thoại hiểu và tìm kiếm đồng thời cả chữ viết, hình ảnh, âm thanh lẫn video một cách nhanh chóng.

Google DeepMind

Bản đồ này dự đoán trước những thay đổi nhỏ trong mã di truyền của con người giống như việc dò trước các lỗi chính tả trong một cuốn sách lớn.

OpenAI

What forward deployed engineering means at OpenAI

Microsoft Research

At a glance Brings rich interactivity to AI-generated charts: Flint-interactive enables agents to create data visualizations that users can explore directly through clicking, filtering, zooming, highlighting, and cross-chart interactions. Makes complex interactions easier

A coding-agent skill for multi-phase security audits with independently verified, machine-readable findings

3k stars · 38.7k total

The PHP Interpreter

423 stars · 41k total

A visual, example-driven guide to Claude Code — from basic concepts to advanced agents, with copy-paste templates that bring immediate value.

370 stars · 42.9k total

Minecraft 启动器 Plain Craft Launcher（PCL）。

115 stars · 8.6k total

FastAPI framework, high performance, easy to learn, fast to code, ready for production

96 stars · 103.6k total

The development of embodied models of animal behavior is a major goal of the neuro-AI community. Embodied models will enhance our ability to make direct comparisons between biological and artificial agents. To make such comparisons effective, though, we need embodied models that

Structural Causal Models (SCMs) provide a principled framework for observational, interventional, and counterfactual reasoning, all of which rely on probabilistic inference. Existing frameworks either transform SCMs into alternative tractable representations or rely on approximat

Watermarking has emerged as a key tool for identifying text generated by large language models (LLMs), with current schemes embedding statistical signals into the generation process to enable later detection. In this work, we identify a shared weakness across a broad class of pop

Instruction-based image editing models can now add, remove a specified number of objects or change their category while preserving the scene. Yet these models often fail to produce the correct number, and these failures grow sharply as scenes become more complex and the requested

Standard LLM-based speech recognition systems process utterances in isolation, limiting their ability to leverage conversational context. Here we study whether multimodal context from prior turns improves LLM-based ASR and how to represent that context efficiently. We find that,

World models that capture how actions induce physical change enable scalable robot learning without reliance on embodiment-specific action labels. Pixel-space video models provide broad visual priors but expend model capacity on dense appearance reconstruction, while direct actio

Prognostic predictions from ML models support human decision making in several fields, including high-stakes ones such as healthcare and the judiciary. Yet, we still lack a clear understanding of how decision makers learn from ML-based decision support (ML-DS). In this paper, we

Activation functions play a central role in neural networks by shaping internal representations. Recently, learning binary activation representations has attracted significant attention due to their advantages in computational and memory efficiency, as well as interpretability. H

For large language models (LLMs), post-training quantization (PTQ) significantly reduces memory and computational overhead. Despite rapid advancements, existing papers often lack experiments on the same ground. In addition, since quantization solutions typically contain multiple

Motivated by ten years of Learning-Based Control Invited Sessions at the IEEE Conference on Decision and Control (CDC), this tutorial paper reviews key advances at the intersection of machine learning and control. We focus on methods that combine the flexibility of machine learni

Diffusion language models generate by iteratively denoising a masked sequence under bidirectional attention, without the strict left-to-right commitment of autoregressive decoding. Whether this difference affects how reliably they remain grounded in structured inputs is largely u

Among existing embedding compression approaches, low-rank factorization provides a simple and mathematically principled solution by approximating the embedding matrix with a reduced-rank decomposition. However, its practical effectiveness is often limited, as the resulting linear

This paper proposes Role-Enhanced Zero-Shot Chain-of-Thought (RE-ZS-CoT), a prompting framework that systematically integrates domain-specific expert roles to enhance reasoning in large language models (LLMs) without task-specific examples. We assess two versions of our approach.

Dependency parsing resource development for Arabic has largely focused on Modern Standard Arabic, leaving dialectal Arabic comparatively under-resourced. We introduce MADARTB, the first large-scale parallel dependency treebank spanning five city dialects: Beirut, Cairo, Doha, Rab

Gold's theorem shows that no learner can identify an arbitrary language in a superfinite class from positive presentations alone. We prove a positive result of a different shape. Fix an innate \emph{recursive contrast operator} $R$ mapping any string to a finite set of structural

This paper explores the potential of Artificial Intelligence (AI) to enable end-to-end automotive design and software development for Software-Defined Vehicles (SDVs). Based on our contributions within the Central Car Server (CeCaS) and follow-up EU-funded projects, we present an

Memorization inflates the benchmark scores of large language models (LLMs), since a model can reproduce answers it saw in training. Researchers test for it with consistent identifier renaming (CIR), which gives every identifier in a task a new name, and treat a lower score on the

The use of AI-based tools in software development is rapidly changing how software is produced, enabling large volumes of code to be generated with limited human implementation effort. However, this raises fundamental challenges for ensuring that generated artifacts are aligned w

Coding agents now complete tasks autonomously, modifying real-world state such as artifacts, software environments, and configurations along the way. Yet they show limited backup awareness: recognizing that an action will modify state that may later be needed, and preserving it b

While large language models are increasingly deployed as tool-calling agents, their safety evaluations still focus primarily on their ability to refuse harmful requests in text. We investigate whether the refusal behavior in text transfers to tool use. We construct a benchmark of

When San Antonio–based software engineer Nikhil Jathar first went to Washington, D.C., to advocate for federal support for technology education and research funding, he was in awe of the opportunity. It was spring of 2025, not long after Jathar settled in the United States from I

IEEE Life Fellow Manfred “Fred” J. Schindler has been elected as the 2027 IEEE president-elect. He will begin serving as president on 1 January 2028.Schindler, who was nominated by the IEEE Board of Directors, received 16,421 votes in the election. Senior Member David Alan Koehle

This White Paper gives robotics researchers and engineers an overview of a new large-scale motion capture dataset built to close the data gap limiting humanoid robot learning. It also shows how policies trained on the dataset transfer to a real humanoid robot.What you will learn

Happy IEEE Day 2026!IEEE Day is an annual event that celebrates the first time in history when engineers worldwide gathered to share their technical ideas in 1884.One of the IEEE Day’s objectives is to show the ways IEEE members, in local communities, join together to collaborate

Các nhà khoa học cho rằng khi máy tính viết tin tức kèm theo cảm xúc để thu hút người đọc, chúng ta cần kiểm tra xem chúng có đang làm con người cãi nhau hay không.

A region of the brain called the striatum is critical for many cognitive and motor functions, including decision-making, control of movement, habit formation, and processing of reward. It also plays a role in addiction and is significantly affected by Huntington’s disease, schizo

When an artificial intelligence image generator produces a portrait, whose work went into it? The question sits at the center of lawsuits, licensing deals, and proposed regulations worldwide. Artists want credit. Companies want clarity. Policymakers want a way to assign responsib

Artificial intelligence models are jacks of many trades, including writing, generating images, and creating 3D models. But they aren’t as helpful when it comes to testing robots or designs for vehicles in diverse environments, since they don’t understand physics as well as they d

Uber is rolling out ChatGPT Work to more than 30,000 employees, helping teams research, analyze data, create presentations, and complete complex workflows.

How OpenAI works with distribution partners to uphold consistent safeguards, support responsible access, and clarify expectations in our Usage Policies.

Generative AI

General Science

Generative AI

At a glance Harnessed Agentic RL: Microsoft Research Asia introduces a training paradigm in which the same agent harness used in deployment participates directly in reinforcement learning, removing the need to reimplement the agent inside the training framework. Lightweigh

The U.S. Department of Commerce has finalized a CHIPS and Science Act R&D award of up to $1 billion for Anderon, a quantum foundry IBM established earlier this year. The 16 September announcement backs a manufacturing service for IBM and other quantum-hardware developers.IBM is m

This article is brought to you by Emerson.I’ve spent much of my career as an engineer, including years in the semiconductor industry. And one lesson has stayed with me through every major technology shift: innovation always creates new complexity.In semiconductors, we have seen t

Supply chains in many industries are full of tangles, but for semiconductor tech, that’s putting it lightly. Many chip components are location-constrained and can easily be throttled by geopolitical factors. Chipmaking is also tied up with social and environmental concerns, and i

Khi các bóng bán dẫn trên vi xử lý trở nên nhỏ đến mức khó tin, việc xếp chồng chúng lên nhau giống như những tầng nhà cao ốc thay vì đặt cạnh nhau giúp giữ cho chip hoạt động mạnh mẽ mà không bị rò rỉ điện.

Các kỹ sư chế tạo chip máy tính đang phải đối mặt với một thách thức lớn khi các linh kiện ngày càng nhỏ đến mức những sai số nhỏ cũng khiến chúng không hoạt động được, đòi hỏi sự phối hợp chặt chẽ từ mọi công đoạn sản xuất như những người thợ ráp hình.

Để hiểu xem một con chip hoạt động tốt đến đâu, các nhà khoa học không chỉ kiểm tra điện áp và nhiệt độ thông thường mà còn phải cho nó chạy các chương trình thực tế giống như một chiếc xe hơi được thử nghiệm trên cả đường băng lẫn đường phố đông đúc.

Công ty này tạo ra loại bộ nhớ máy tính mới lưu giữ dữ liệu cực nhanh và bền bỉ để dùng trong máy tuốt lúa dữ liệu lớn.

Các kỹ sư chế tạo một chiếc la bàn phụ dùng sóng điện thoại để tìm đường khi bản đồ vệ tinh bị kẻ xấu làm nhiễu.

Công ty này kiểm tra thật kỹ các linh kiện nhớ máy tính giống như thử nghiệm sức chịu đựng của đồ chơi trước khi bán cho mọi người.

Mixture-of-Experts (MoE) decoding on scratchpad-based tensor accelerators (STA) is dominated by moving expert weights while the compute engines sit idle. This traffic is hard to hide, because the experts are known only after routing, and hard to shrink without losing quality or a

Syndrome extraction for a CSS code has depth at least the Tanner-graph degree $\Delta$, attained by edge colouring if gate order is ignored; correctness adds a parity constraint. We prove three general obstructions to depth-$\Delta$ extraction and show that, for every two-block g

Speculative decoding accelerates LLM inference, but deploying it on CPUs introduces severe systems bottlenecks: context-swapping latency between draft and target graphs, and memory allocation overhead during KV cache rollbacks. We introduce SubTernary, an execution framework that

The Open Compute Project's microscaling (MX) family includes MXFP4, which stores 32 four-bit elements with a shared power-of-two scale. When lowering a dot-product reduction, a compiler can apply these scales before the reduction, or defer scaling until afterwards. Although equiv

Throughout her career, roboticist Barbara Mazzolai has turned to nature for inspiration. Now she wants to ensure the technology she builds gives back to the environment, too.After starting her career as a biologist, a chance opportunity saw Mazzolai switch streams to engineering

You sit down and put your arm in the cradle. You press a button. The machine takes it from there.A near-infrared light sweeps your inner elbow, hunting for a vein. A puff of alcohol hits your skin. An ultrasound probe glides across your arm, mapping how deep the vessel runs and w

Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion.Humanoids Summit Seoul: 22–23 Septem

Recently, multi-node inertial measurement unit (multi-IMU) systems have shown promise for robust odometry in legged robots. Multi-IMU fusion can improve legged robot odometry but requires accurate spatiotemporal calibration among the sensors. Existing multi-IMU calibration method

Ultrasound examinations typically follow specific scanning protocols that require navigating through an ordered sequence of anatomical landmarks and target imaging planes. We present a local action-conditioned world model framework for goal plane probe guidance in robotic ultraso

Vision-based manipulation policies are commonly learned from rich visual observations, typically using dense visual features extracted from full resolution images. It remains unclear how much of this visual information is really needed for effective policy learning. In this paper

Recent world action models provide visual foresight and future prediction for robot manipulation, but lack explicit mechanisms for compositional skill reuse. Agentic robotics combines code-based planning and reasoning with reusable skills, but often lacks predictive grounding in

For ten days, auditors appointed by the Korea Internet & Security Agency (KISA) sat together with us and worked through 80 controls. That is what it took for a company headquartered in the United States, operating a global network serving Korean users, to be certified in South Ko

AI can write code and pass the bar exam, but it still struggles to buy a pair of shoes online.That’s because websites were built for humans, not agents. Most agents that shop for someone today work from a site’s HTML — they look at the page, guess which button does what, click, a

As fun as AI is, we built so much bad software over the last 30 years, and now the universe is punishing us for our sins. — Armin Ronacher Tags: security, generative-ai, armin-ronacher, ai, ai-security-research, llms

Today's PHP Security Release, and How We're Strengthening PHP Security If you're running anything on PHP it's vital you upgrade - today's updates (across the entire PHP 8.x line) address two remote code execution holes, and there have been exploits circulating since last night, t

Research: sparqlite: SPARQL 1.1 inside SQLite, written in Rust This research task is really an experiment in running Claude Code for web in "Ultracode" mode, as described in Orchestrate subagents at scale with dynamic workflows in the Claude Code documentation. With Ultra

The concept of recursive self-improvement (RSI) dates back to I. J. Good (1965), where he defined an “ultraintelligent machine” as a system that can surpass humans in all intellectual activities and design better machines to improve itself. Yudkowsky (2008) used the phrase “recur

Scaling laws are one of the most critical empirical findings in deep learning. The observation is simple in form: the training loss $L$ decreases predictably as we scale up model size $N$, dataset size $D$, and compute $C$, following a power-law curve, which appears as a straight

104 likes

100 likes

662 points

645 points

595 points

567 points

552 points

511 points

515 points

486 points

424 points

396 points

132 upvotes

68 upvotes

64 upvotes

44 upvotes

43 upvotes

47 upvotes

32 upvotes

26 upvotes

Công cụ mới giúp các chương trình viết bằng Java tận dụng trực tiếp sức mạnh siêu tốc của card đồ họa, giống như đổi từ đi xe đạp sang đi tàu điện cao tốc.

Bài viết giải thích cách một hệ thống cơ sở dữ liệu lưu lại ảnh chụp nhanh dữ liệu mà không làm gián đoạn công việc đang chạy, giống như chụp lại bức ảnh của một người đang chạy mà không bắt họ phải đứng yên.

Bài viết vạch trần những bài kiểm tra tốc độ giả mạo của một ngôn ngữ lập trình mới, giống như việc gian lận trong cuộc thi chạy để nhận huy chương oan.

Nhà phát hành công bố bản vá lỗi bảo mật quan trọng cho ngôn ngữ PHP nhằm giúp các trang web tránh bị kẻ xấu xâm nhập giống như việc thay ổ khóa mới chắc chắn hơn cho ngôi nhà.

Đây là một bộ sưu tập công cụ giúp ngôn ngữ Java có thể mở ra hầu hết mọi loại tệp nén khác nhau giống như một chiếc chìa khóa vạn năng mở được mọi loại khóa.

Công cụ này giúp nhóm lập trình viên tạo các đường dẫn ngắn gọn ngay trong kho mã nguồn để dễ dàng chia sẻ tài liệu với nhau giống như tạo những con đường tắt trong một khu vườn lớn.

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady ...

51 reactions

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady 💡...

45 reactions

I've had this conversation more times than I'd like to admit. I raise a technical question in a...

39 reactions

One click on "Place order." Two rows in the orders table, created 40 milliseconds apart, identical...

37 reactions

I spent a ridiculous amount of time trying to get WebGPU working properly on my Linux machine. I...

32 reactions

I can generate a full authentication flow with a database schema, API routes, and unit tests in under...

23 reactions

The ancient city of Qianzhou on the bank of the Wanrong River has a long history and is known as the first of the four ancient towns in Xiangxi. Today, the ancient city of Qianzhou is not only a cultural window of Xiangxi Prefecture, but also a gathering place for the entire inta

The project is located in the west of Xike Netease Science and Technology Park, Binjiang District, Hangzhou City, Zhejiang Province, just one way from Alibaba. The whole building is divided into three parts in a vertical direction, with 1 floor as a shared public service space, 2

Beijing's neighborhood elderly apartment is designed to redefine the concept of an old-age apartment in order to change the condition of old and poor, which is 4,500 square meters.

Situated in a rural Boulder neighborhood, Parkside Villa is a comprehensive transformation of an existing 9,100-square-foot residence, reimagined for the evolving lifestyle of an active family. Rather than demolishing and rebuilding, Studio B Architecture + Interiors approached t

The Faculty of Wood Engineering and Creative Industries building is located in the Botanical Garden of the University of Sopron. The design prioritizes the relationship between old and new, between built space and the natural environment, and the practical, flexible character tha

VISION COMPETITION EXHIBITIONThe Vision Competition Exhibition, entitled “From Void to Value: Revisioning Tallinn’s Old Town”, is presented at the Tammsaare Park Outdoor Gallery in Tallinn opening on 16 October and running until 12 November 2026. The exhibition presents proposals

What does it take to design a capital city for a culturally diverse nation from the ground up? This book offers a rare, behind-the-scenes look at Nusantara, the new capital of Indonesia, told through the eyes of the lead designer from its earllest conception to its current develo

The open call for the 19th Istanbul Biennial aims to deepen a particular line of research within the Biennial’s broader conceptual framework. In their invitation text to artists, the curators state: “Art cannot end wars or resolve their underlying causes, but it can open spaces f

In 1972, Robert Venturi, Denise Scott Brown, and Steven Izenour published Learning from Las Vegas, challenging architects to reconsider the underappreciated, largely dismissed commercial landscape in the United States. Their storied distinction between the "duck," a building whos

Aalok Farmhouse is a 3-bedroom weekend home located on the outskirts of Bengaluru that reinterprets the farmhouse typology through a series of independent pavilions connected by landscape. Rather than treating the outdoors as an extension of the built form, the project allows cou

About an hour's drive from New York, the Corwin House sits in hilly woodland high above the Saugatuck River. Completed by Richard Neutra in 1956, it is one of only three houses the architect built in Connecticut. Architectural designer Aziz Osko has updated the house, revisiting

Các kiến trúc sư đã sửa sang và nối dài một ngôi nhà đá cổ ở Anh bằng một phần nhà mới gắn đồng hồ mặt trời.

Bảo tàng về ban nhạc The Beatles ở Anh sẽ chuyển đến một khu mua sắm cũ từ thập niên 60 và biến tháp phát thanh thành đài quan sát hình tàu ngầm vàng.

Các kiến trúc sư đã khéo léo xây một ngôi nhà hẹp trong hẻm ở Việt Nam với nhiều khoảng sân nhỏ và mái ngói dốc để mang lại cảm giác thoáng đãng.

Chính phủ Albania bị phàn nàn vì vội vã duyệt xây khu nghỉ dưỡng sang trọng có liên quan đến gia đình Trump trên vùng đất thiên nhiên được bảo vệ.

Bài viết giới thiệu bảy công trình kiến trúc nổi bật có sử dụng tông màu đỏ tươi sáng.

the loneliest plant on earth lives under glass   Inside London’s Kew Gardens, a cycad called Encephalartos woodii has spent more than a century beneath the glass roof of the Victorian Temperate House. Its enormous crown of dark green fronds appears perfectly healthy, though it ma

CORTEN STEEL CONDUITS FRAME REVERSIBLE UNDERGROUND structure   Positioned at an altitude of 2,173 meters at the Pavillon intermediate station of the Skyway Monte Bianco cableway in Aosta Valley, Italy, Tana delle Marmotte (Marmot Den) is a permanent underground educational instal

AA Apartments reuse open-plan commercial space in Málaga   Designed by CREARCTIVE, AA Apartments addresses the adaptive reuse of a former open-plan commercial space in Málaga, converting the premises into four independent residential units. To make full use of the building’s four

SUBTERRANEAN ART COMPLEX EMBEDDED IN BURGAS EXCAVATION VOID   Designed by UNFORMED DESIGN, Ground of Descent won first prize among 145 international entries in the Park Zone 3 competition organized by the Municipality of Burgas and the Burgas 2032 Foundation. The proposal transfo