Our whitepaper on Isaac Lab is out! Isaac Lab is a natural successor of Isaac Gym that pioneered GPU-accelerated simulation for robotics. It subsumes all the features of Gym and provides the latest advances in simulation technology to robotics researchers. It also supports

🔷 Introducing Rig3R - our new geometric foundation model for 3D perception in AVs. wayve.ai/thinking/rig3r/

Tesla - collects 4.3M hours of driving data - every day - for free - to train a 2DoF system (steering + throttle). - yet full autonomy remains unsolved. Frontier robotics startups/labs - collect or purchase 0.01M–1M hours of data - every X month - for millions of dollars - to

We release OpenReal2Sim, an open-source toolbox for real-to-sim reconstruction and robot simulation. A key difference from prior work is our focus on building an interactive digital twin from in-the-wild data — even Internet images or generated videos. Try it out: Interactive

Build smarter robots, faster. 🤖 With NVIDIA Isaac GR00T-Dreams, built on NVIDIA Cosmos, developers can generate unlimited synthetic data from a single image and natural language, or generate data from “lucid dreams” via teleoperation for more complex tasks. See it in action.

Zero teleoperation. Zero real-world data. ➔ Autonomous humanoid loco-manipulation in reality. Introducing VIRAL: Visual Sim-to-Real at Scale. We achieved 54 autonomous cycles (walk, stand, place, pick, turn) using a simple recipe: 1. RL 2. Simulation 3. GPUs Website:

Researchers are exploring Marble’s generative 3D worlds as a way to rapidly produce simulation-ready environments for robotics without manual scene construction.

nano banana pro can perform arbitrary transformations of sets of images into new images. it's an LLM but for pixels and the first image model that isn't a toy. many of you are sleeping on this, but if you have any experience with 3d graphics the images below may help you wake up.

🚨Google and UCSD just introduced Radiance Meshes, a new radiance field representation that produces watertight meshes and renders faster than 3DGS. Code and demos are available now. Code: github.com/half-potato/ra… Demos: half-potato.gitlab.io/rm/#demos

State of Robot Learning - Dec 2025 In this blogpost I lay out: - how robot learning is done today (behavior cloning), focused on problem formulation and data sourcing - why we don't currently use other approaches (e.g. RL) - predictions for the future + startup advice (1/N)

Most past work throws human data into a pretraining mix. EgoMimic showed that, with proper alignment, you can co-train with human data. In his internship project at Pi, Simar Kareer took this a step further and showed that human data can "post-train" VLAs. This enables robots

Beginning of an exciting journey! 🤖🤩

微软新3D建模大模型 Trellis.2 实测 微软刚刚发布了新模型 trellis 2, 4B 这是一个通过图片就能创建3D模型的大模型, 它使用了一个叫"基于稀疏体素的 3D VAE 流匹配变换器"的新方法, 从官网看效果很惊艳, 但是打开它的demo, 就能发现一些问题, 模型有孔洞, 可以看这个官方demo,

9M+ downloads worldwide. 🎉 In 2025, NVIDIA’s open robotics datasets topped the charts. Leading the way was the GR00T post-training dataset, @HuggingFace’s most downloaded robotics dataset with 835K downloads last month. Helping robots learn faster, everywhere. 🦾 Read the

Excited to share Large Video Planner (LVP) -- a open source video-based robot foundation model trained Kempner Institute at Harvard University that can zero-shot generalize across both domains and robots. Through third-party evals, LVP outperforms both SOTA VLAs and video models across novel tasks/robots!

Can we train large world models in 3D? PointWorld is an interactive (action-conditioned) 3D point-cloud world model for robotics. Trained on large-scale data, PointWorld can generalize to novel RGB-D images and solve new real-robot tasks (no demos required) Check it out :)

We release Action100M, the hero behind VL-JEPA. It is a large dataset with O(100 million) dense action annotations on HowTo100M procedural videos. We hope it serves as a robust data foundation to advance physical world modeling research.

Rethinking Video Generation Model for the Embodied World

Thrilled to launch Project Genie, an experimental prototype of the world's most advanced world model. Create entire playable worlds to explore in real-time just from a simple text prompt - kind of mindblowing really! Available to Ultra subs in the US for now - have fun exploring!

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…