Zhiding Yu (@ZhidingYu) profile picture

Working to make machines understand the world like human beings. Words are my own.

Santa Clara
Joined July 2020
Embodied models don't just think in text, they can think in visual plans, and fast. Introducing VFlash, System-1 VLM for Physical AI. Joint work by @YuCheng_Chou_, Shihao Wang and Jin Huang, LPR Team (@jankautz), GEAR Lab (@DrJimFan), and many other collaborators. 🌐 Website: research.nvidia.com/labs/lpr… 🧵[1/8] More details in the thread👇
5
28
7
141
25,090
Super cool.
Dear researchers: Did you know that your H100 can actually ray-trace like a gaming card? 🎮 for a long time, people have assumed that compute cards like the H100 can only compute, and that you need a graphics card like the RTX 5090 to render. we show that's wrong. 🚨Meet RTX H100. Blender Cycles on the Open Data benchmark (Blender 5.3.0): ours 48,641. the RTX 5090 is at 19,202, and a stock H100 gets only 7,041. our original goal was to match the 5090. we ended up at 2.5x. our neural rendering approach is lossy. we test within a reasonable loss, and every result passes an image check. we think nobody doing this before was an accident: graphics people rarely touch compute cards, and people who have compute cards rarely know graphics. robotic simulators need fast rendering. now a general-purpose compute card can do it, so you don't need heterogeneous compute or a complicated cluster. we'll extend this to more hardware next. almost ready for GTA 6 on the cluster 🤣 Best, Wenhao
1
9
1,609
Zhiding Yu retweeted
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me. I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA. I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation. Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO. Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours. Jensen
60,962
121,707
24,488
1,173,906
58,569,541
Disgusted by what passes for "journalism" at WSJ, NYT, and The Economist. The internet has a long memory. Laundering competitor smears and unverified rumors as "national security" won't age well. When the dust settles, history will remember who stood for facts—and who orchestrated a hit job.
From @WSJopinion: Jensen Huang’s resignation from Nvidia was necessary, but it also does regrettable harm to the company and America’s AI industry. Maybe that was China’s plan. on.wsj.com/4k9p4V2
29
47
4
466
15,069
Replying to @WSJ
I can't believe this garbage. The article starts out by saying that NVIDIA's export strategy was lawful, then ignores that and tears into the company and its lawful export strategy with unsourced unsubstantiated suggestions that other companies might have done bad things.
57
162
9
3,231
56,948
I'm very sad to see Jensen go. One of Jensen's big legacies to the AI community is the call for open-sourcing and healthy competition. From now on, I will double down on this effort. "Open source versus closed source. Light versus darkness. A decisive battle between two futures and two destinies for humanity."
Back in 2023, when I was doing my PhD, I received two internship offers: one from Microsoft and the other from NVIDIA. I chose NVIDIA for my first and only internship because of the incredible things Jensen and his team had achieved. I've always admired Jensen's leadership, and I still believe he is the soul of NVIDIA. But what's happening now is deeply disappointing. Perhaps it's also bad for the open-source community.
7
23
2
267
10,867
Zhiding Yu retweeted
Introducing GR00T 2: our most advanced, open System One model for robotics. It runs entirely on your robot’s onboard compute, fully untethered, and you own every weight of it. GR00T 2 features: 1. Long-horizon tasks: dexterous manipulation and whole-body control chained over 20+ steps and 5+ minutes, such as performing chemistry lab experiments or assembling model airplanes end to end. 2. X-Adapt: adapting from easy data sources (human wearables and simulation) and deploying zero-shot to diverse real robots. 3. Unified Cognition: native integrations with frontier agents for advanced physical reasoning, robust error recovery, and safety monitoring. It's an open-source buffet! We are releasing not only the model weights and X-Adapt recipes (coming December), but also our agentic framework, simulation engine, annotation engine, hardware design, and even training data (100K hours, that's ~11 yrs of experience), all available today! Everything that moves will be autonomous. Anyone can become a roboticist. Lock in, build with us! Deep dive in thread:
195
267
101
1,964
185,550
Zhiding Yu retweeted
Amazing work by the entire NVIDIA robotics team!
A finer touch for humanoid robots. 🤖 At #CoRL2026, NVIDIA is introducing GR00T 2, an open world action model for humanoids, alongside new open industrial benchmarks from @RLWRLD_ai and @UCBerkeley built on Isaac Lab-Arena. Explore how NVIDIA’s open world action models, scalable simulation tools and agentic workflows help researchers train, test and refine robot skills, from connecting cables to assembling small components. Discover more ➡️ nvda.ws/4ixvdZV
3
18
1,173
GR00T 2 is a giant leap in System-1 robot foundation models. Very glad to be part of this effort!
Meet GR00T 2, our new generation of world-action models (WAMs): it imagines how the physical world will unfold and generates the actions that get it there. One model runs on parallel grippers, dexterous hands, and humanoids. research.nvidia.com/labs/gea…
1
1
35
2,311
A finer touch for humanoid robots. 🤖 At #CoRL2026, NVIDIA is introducing GR00T 2, an open world action model for humanoids, alongside new open industrial benchmarks from @RLWRLD_ai and @UCBerkeley built on Isaac Lab-Arena. Explore how NVIDIA’s open world action models, scalable simulation tools and agentic workflows help researchers train, test and refine robot skills, from connecting cables to assembling small components. Discover more ➡️ nvda.ws/4ixvdZV
69
89
61
614
73,264
Zhiding Yu retweeted
YouTube live streaming link, 1:45 pm! youtube.com/live/F_4AAOQsWlA
CoRL Keynote. Hilton Austin. Monday Nov 9, 1:45 pm. 2 days countdown. Something extraordinary.
16
17
6
204
37,523
Zhiding Yu retweeted
i went to bed and asked claude to paint an autumn tree overnight. at around 4am, it realised that it didn't have some shades -- so it stopped and titrated its own paint: diluting red, sienna and madder step by step on a second palette, then measuring each pocket from a photo. woke up to this beautiful watercolor painting. 303 brush marks, took about 4.5 hours 🍂 (still working on the speed)
508
750
530
9,416
2,852,386
Zhiding Yu retweeted
CoRL Keynote. Hilton Austin. Monday Nov 9, 1:45 pm. 2 days countdown. Something extraordinary.
80
121
64
1,358
240,985
VFlash code has been released. Welcome to try out! github.com/NVlabs/Eagle/tree…
Embodied models don't just think in text, they can think in visual plans, and fast. Introducing VFlash, System-1 VLM for Physical AI. Joint work by @YuCheng_Chou_, Shihao Wang and Jin Huang, LPR Team (@jankautz), GEAR Lab (@DrJimFan), and many other collaborators. 🌐 Website: research.nvidia.com/labs/lpr… 🧵[1/8] More details in the thread👇
1
19
1,103
具身里面重要的不是System-1 vs System-2的对比,而是能力边界的划定。什么能力适合System-1并行输出,什么能力需要System-2的自回归输出,这才是接下来的重点。VFlash只是一个开始。
Replying to @ZhidingYu
@ZhidingYu 老师这波把 System-1 做成了可执行的视觉规划,绕开了文本规划那段中转。具身里 System-1 和 System-2 讨论了很久,真正稀缺的是延迟与可执行性同时过关。等你们的数字。
2
6
46
3,103
Zhiding Yu retweeted
REINFORCE Zero: Trainer–Inference Invariance (TII) + the simplest IS ratio keeps async LLM RL stable for 20,000 updates. To our knowledge, the first open-source 20K demo of this recipe: unclipped token IS, no rejection gates, a minimal loss.
16
33
9
357
50,459
Thank you Jim and GEAR Team for the great support on this work along the way. A lot of exciting opportunities in System-1 models looking ahead!
The era of System One Vision Model is here! Introducing VFlash: parsing the visual world in parallel at unmatched speeds for Physical AI. Most embodied prediction tasks like object localization, robot trajectory planning, and dense 3D motion are fixed-length float vectors. For example, a bounding box is always 6 tokens, a 10-waypoint robot arm trajectory is 54: we know these beforehand. Instead of decoding one token at a time, we can output the entire geometric shape in one go. VFlash enables up to 12x faster action decoding and a 5.1x overall robot inference speedup on edge GPUs like the Jetson Thor and DGX Spark. What’s surprising is that it ends up even more accurate than the VLM backbone, with up to 61% lower 3D trajectory error. The intuition is that VFlash avoids the cascading errors of autoregressive decoding, where every coordinate is written in ink and every later point loses the chance to backspace. Everything that moves will be autonomous, and everything autonomous will need to think on its own chip. The next inference revolution is physical reasoning at the speed of reflex. And as always, it’s open paper, open source! Link in thread:
10
992
🧵[2/8] What is a System-1 VLM? VLMs are often used to express visual plans in terms of bounding boxes, points, motion trajectories, and more in physical AI applications. In these cases, the outputs are autoregressively decoded in a sequential way. VFlash is a generalist Physical AI decoding method that turns sequential decoding into parallel ones, fast and accurate. Similar to a JEV model, VFlash enables VLMs with System-1 like reactive capabilities in visual planning.
Embodied models don't just think in text, they can think in visual plans, and fast. Introducing VFlash, System-1 VLM for Physical AI. Joint work by @YuCheng_Chou_, Shihao Wang and Jin Huang, LPR Team (@jankautz), GEAR Lab (@DrJimFan), and many other collaborators. 🌐 Website: research.nvidia.com/labs/lpr… 🧵[1/8] More details in the thread👇
4
7
48
4,137
🧵[7/8] Performance on NVIDIA stack We supported inference acceleration all the way to deployment on Jetson Thor, DGX Spark and H100. Built on NVIDIA’s TensorRT-Edge-LLM and TensorRT inference stack, VFlash brings faster visual planning to both robot manipulation and autonomous driving. On Jetson Thor and DGX Spark, VFlash delivers 5× faster inference on MolmoSpace, over 3× faster inference on NAVSim, and up to 5.7× faster inference on Waymo E2E compared with AR decoding baselines.
1
1
2
238
🧵[8/8] Team work We built VFlash to make visual planning fast, accurate, and useful for action, taking a step toward System-1 VLMs for physical AI applications. This was a team effort. Check all the contributors and explore the demos / full results at below: 📄 Technical report: research.nvidia.com/labs/lpr… 🌐 Project: research.nvidia.com/labs/lpr… 💻 Code: github.com/NVlabs/Eagle/tree… Thank you to all who made VFlash possible.
1
2
174