In weekly meeting, our advisor Huazhe kept asking: "What's the real point of simulation for manipulation?" HERMES offers a potential answer. Our framework converts diverse human motion data into real robot behaviors via sim training, all without task-specific reward designing.

Very impressive！Focused sim2real mobile dexterous manipulation，human data is leveraged for the guidance in reinforcement learning.

solid deployment🫡

We just did World’s first on-stage autonomous demo of long-horizon dexterous VLA 🚨 No training. No setup. Performance out of the box. Live demo is hard and unpredictable, but we felt great about our model’s generalization, and it went pretty well! 💯 Zero-shot. 100% success.

Nice work! Our recent project HERMES also leverages a similar pipeline for bimanual dexterous manipulation — using a unified reward and egocentric depth-based sim2real to deploy in diverse in-the-wild scenarios for a variety of tasks. gemcollector.github.io/HERMES/

Come by to the MuJoCo meetup at CoRL 🇰🇷. I'll be there if you want to chat! github.com/google-deepmin…

Meet mjlab. Powered by MuJoCo Warp. Drops Monday.

Our grand finale: A complex, long-horizon dynamic sequence, all driven by a proprioceptive-only policy (no vision/LIDAR)! In this task, the robot carries a chair to a platform, uses it as a step to climb up, then leaps off and performs a parkour-style roll to absorb the landing.

🚀 Introducing SARM: Stage-Aware Reward Modeling for Long-Horizon Robot Manipulation Robots struggle with tasks like folding a crumpled T-shirt—long, contact-rich, and hard to label. We propose a scalable reward modeling framework to fix that. 1/n

cool! Can’t wait to try it!

We open-sourced the full pipeline! Data conversion from MimicKit, training recipe, pretrained checkpoint, and deployment instructions. Train your own spin kick with mjlab: github.com/mujocolab/g1_s…

three years ago, DiT replaced the legacy unet with a transformer-based denoising backbone. we knew the bulky VAEs would be the next to go -- we just waited until we could do it right. today, we introduce Representation Autoencoders (RAE). >> Retire VAEs. Use RAEs. 👇(1/n)

Imitation learning provides a solid prior, and Online learning further refines the policy for better performance. In the lab, I’ve been watching Kun’s policy grow stronger over time.😆

Introducing Huazhe Harry Xu HERMES, a unified human-to-robot learning framework built on the Galaxea mobile base and A1 dual-arm platform. With high-fidelity simulation and dexterous hands, HERMES enables robust sim2real transfer for complex mobile manipulation tasks. #Robotics

Excited to share SoftMimic -- a new approach for learning compliant humanoid policies that interact gently with the world.

Ever want to enjoy all the privileged information in sim while seamlessly transferring to the real world? How can we correct policy mistakes after deployment? 👉Introducing GSWorld, a real2sim2real photo-realistic simulator with interaction physics with fully open-sourced code.

Unified multimodal models can generate text and images, but can they truly reason across modalities? 🎨 Introducing ROVER, the first benchmark that evaluates reciprocal cross-modal reasoning in unified models, the next frontier of omnimodal intelligence. 🌐 Project:

After a year of team work, we're thrilled to introduce Depth Anything 3 (DA3)! 🚀 Aiming for human-like spatial perception, DA3 extends monocular depth estimation to any-view scenarios, including single images, multi-view images, and video. In pursuit of minimal modeling, DA3

Introducing Gallant: Voxel Grid-based Humanoid Locomotion and Local-navigation across 3D Constrained Terrains 🤖 Project page: gallantloco.github.io Arxiv: arxiv.org/abs/2511.14625 Gallant is, to our knowledge, the first system to run a single policy that handles full-space

Today, we present a step-change in robotic AI Sunday. Introducing ACT-1: A frontier robot foundation model trained on zero robot data. - Ultra long-horizon tasks - Zero-shot generalization - Advanced dexterity 🧵->