Independent global robotics news from Article module 102.
Each record is classified by both destination market and product layer, including complete robots and core components.
At the CeMAT site, in order to demonstrate the ability to collaborate in depth with multiple products, PUDU DUDU robots converted the stand into a real in-store operation...
DiT, U-Net, and latent spaces are three important concepts behind modern generative models. This explainer compares their roles in image and video generation and clarifie...
Daimon Robotics is building tactile intelligence around both data collection and model design. The article explains how tactile sensors, datasets, and world models can be...
Yesterday, she sent a farewell letter and jokingly told netizens, "Wait until you get to AGI", and everyone thought she was going to start a business. As a result! Today!...
A founder born in 1998 and educated at Tsinghua has built a world-model startup that has raised about RMB 700 million. The article looks at the company’s technical direct...
However, language is essentially an orderly Token sequence, and vision is a continuous high-dimensional distribution, both of which have a natural gap in information stru...
With the length of the task chain, the computing costs of each step of the model, the long-text capability, affect the end-user experience and cost, the importance of the...
Taobao’s MNN team has open-sourced a 0.3B model optimized for on-device inference. The model reportedly runs offline at up to 350 tokens/s on an iPhone 18 Pro, illustrati...
Pinduoduo’s automated strawberry-farm experiment uses sensors, environmental controls, and software to manage a compact growing facility. After 170 days, the project offe...
Anthropic’s Claude Code organization has seen a series of senior departures, including product leader Cat Wu, who has joined OpenAI. The personnel moves offer a glimpse i...
A widely discussed MIT study claimed that Astra could substantially improve Qwen’s performance by editing a tiny fraction of its weights without conventional training. Re...
Chang'an: PhD student at the School of Artificial Intelligence at Tsinghua University, researching game intelligence. Suey: PhD student at MMLab at the University of Hong...
This is the most noticeable change to ChatGPT Images 2.6 released by OpenAI on November 4th. The new version supports direct GIF generation and can continue to modify ima...
Human beings need to represent the shape and direction of an object in their minds, understand the relationship between different points of view, and infer three-dimensio...
In the summer of 2026, AI volunteered to completely fire out the circle. Thousands of questions, bean bags, Kimi rounds up, and all the big models hated to swallow the an...
In 2026, Big Model sold Token, a business that has been unimaginable. At this year's I/O conference, Google released a figure: 32 trillion tokens were processed each mont...
Human-powered robots are moving from a "power-driven" demo to a "power-driven" commercial phase, where the real mass-production rhythm is stuck, often not with large mode...
In the last few decades, the group has become a global leader in scale, cost and delivery efficiency. It produces more than half of the world's computers, is the world's ...
Shanghai AI Lab’s RoboAtom introduces fine-grained temporal supervision for robot trajectories, helping models learn what happens at each step of a task. The proposed Phy...
Claude Code has acknowledged that its agent can repeatedly reread a full conversation while updating terminal status, consuming tokens for a very small visible change. Th...