David Held

Excellent analysis by @Majumdar_Ani, recommended read!

@siddancha What is the consumer business case for this? Do you mean for some kind of robot sports competition? Or as a training partner for athletes?

OMG it's finally here! Humanoid lawn tennis!!! 🤖+🎾 zzk273.github.io/LATENT/ I have a feeling that robot tennis/table-tennis will be the first successful consumer application of humanoid robots even before they start doing household chores.

🚀 #ICLR2026 Oral 💥 How can we design world models that capture object interactions directly from pixels? Introducing Latent Particle World Models-the first end-to-end self-supervised, object-centric world model, trained from videos, supporting action/img/lang conditioning. 1/n

@bowenwen_me Looking forward to it! I would also love to see a comparison to MapAnything or DepthAnything3.

Our paper has been accepted to CVPR 2026🎊. Code will be released very soon! Stay tuned at github.com/NVlabs/Fast-Fo…

@m_wulfmeier Robot vacuuming!

Name #robotics use cases where 90% success rate is enough, go!

@bowenwen_me Any guesses when you will release the code for this method?

A new milestone for real-time accurate 3D spatial computing! Introducing ⚡️Fast-FoundationStereo⚡️, a real-time zero-shot stereo depth estimation model that accelerates the original FoundationStereo by >10x with comparable quality. Details in threads 🧵 (1/N)

@vincesitzmann Clarification: you call a point cloud an "intermediate representation" but what about 3D sensors that directly record point clouds? Should we not develop methods that can learn from such data?

In my recent blog post, I argue that "vision" is only well-defined as part of perception-action loops, and that the conventional view of computer vision - mapping imagery to intermediate representations (3D, flow, segmentation...) is about to go away. vincentsitzmann.com/blog/bitter_le…

@xwang_lk @jon_barron @TuliMathieu So "natural" just means "recorded using sensors similar to the sensors that humans have"? What about sonar (bats)? And why couldn't some alien species have a lidar sensor built in, making it "natural"?

@jon_barron @TuliMathieu Agreed

About a year ago, I had a discussion with several 3D researchers around a view I still hold today: 3D representations are human-engineered, non-natural abstractions. Using explicit 3D representations as an intermediate step to enhance spatial reasoning in large models may be a detour. A more direct path is to learn from the most fundamental signal we have — raw video data itself. That said, 3D research is absolutely valuable. It plays a critical role in domains where 3D representation is the end product, such as game development, digital content creation, and 3D printing. The question is less about whether 3D matters, and more about what the ultimate objective is.

@nblqbl Now try this with humans.

Here is a game: ask an AI to generate an image. Ask a fresh model to describe it. Use that description to generate a new image. Repeat and see what happens. The results are strange and unsettling. To start: "An apple sits on a table." 1/N

@bowenwen_me Great work! How does it compare to MapAnything or Depth Anything 3?

Our code, pretrained weights, and pseudo-labeled internet stereo data will be released. Welcome to watch and star our repo! (8/N) Code: github.com/NVlabs/Fast-Fo… Webpage: nvlabs.github.io/Fast-Foundatio… Paper: arxiv.org/abs/2512.11130

@AvivTamar1 What is the max score here? What if I want to say that their ability is much greater than is reflected by their achievements?

@davheld The only Nash equilibrium in this game is to give max score on all questions, no?

I am filling out a fellowship application letter of recommendation, and one question is: "Applicant's achievements reflect his or her ability: * - select -Not at All; Well; Moderately Well; Very Well; Extremely Well". If I select "Well" then what does that imply?

@AvivTamar1 What would a person mean if they said it?

Discussing a research idea with ChatGPT. Response I got began with "Love this line of work—..." What does "love" mean here exactly?? So confusing.

What are the best slides you have seen to explain transformers to someone who has never seen them before?

@_wenlixiao Where can I find the paper?

What if robots could improve themselves by learning from their own failures in the real-world? Introducing 𝗣𝗟𝗗 (𝗣𝗿𝗼𝗯𝗲, 𝗟𝗲𝗮𝗿𝗻, 𝗗𝗶𝘀𝘁𝗶𝗹𝗹) — a recipe that enables Vision-Language-Action (VLA) models to self-improve for high-precision manipulation tasks. PLD couples real-world residual reinforcement learning with standard supervised fine-tuning — letting robots discover, recover, and distill their own data flywheel. Quick 🧵

Introducing FMVP: a method that adapts to natural arm motions during robot-assisted dressing. Pre-trained on vision in sim, fine-tuned with limited real-world vision+force data, and tested in a 12-user, 264-trial study, FMVP is robust across garments and motions. #CoRL2025

Check out our new CoRL paper on how to find shapes that cause failures for robot manipulation policies!

Check out our CoRL 2025 paper on learning from past interactions; we show, theoretically and experimentally, the benefit of using a generator-verifier paradigm!

Check out our new CoRL paper on planning in continuous action spaces by searching over sequences of object transformations!

@Nik__V__ Great work!!! Where can I find inference speeds?

Meet MapAnything – a transformer that directly regresses factored metric 3D scene geometry (from images, calibration, poses, or depth) in an end-to-end way. No pipelines, no extra stages. Just 3D geometry & cameras, straight from any type of input, delivering new state-of-the-art results 🚀 One universal model enables SoTA for: 🔥 Mono Depth Estimation 🔥 Multi-View SfM 🔥 Multi-View Stereo 🔥 Depth Completion 🔥 Registration … and many more possibilities! – plus everything is metric 🎯 We release code for data processing, training, benchmarking & ablations – everything Apache 2.0! Details & Links 👇