Chenhao Li
906 posts
@breadli428
Student Researcher @GoogleDeepMind | Embodied intelligence and robot learning | Doctoral fellow @ETH_AI_Center, @leggedrobotics | Prev. @MIT, @ETH_en, @MPI_IS.
Introducing ✨RigidFormer: Learning Rigid Dynamics with Transformers - our attempt to scale learning-based physical dynamics with Transformers. RigidFormer learns rigid dynamics with Transformers. It is a mesh-free, object-centric Transformer for multi-object rigid-body contact dynamics from point clouds. Learning physics with purely neural simulators, without relying on traditional physics engines, is an important and widely studied problem. Prior SOTA methods often use graph neural networks for accuracy and generalization, but still struggle with efficient, high-fidelity simulation at scale. RigidFormer uses only point inputs, matches or outperforms mesh-based baselines on standard benchmarks, runs much faster, generalizes across point resolutions and datasets, and scales to 200+ objects. We also show a preliminary extension to command-conditioned articulated bodies by treating body parts as interacting object-level components. RigidFormer is mesh-free: it does not require mesh connectivity, SDFs, or vertex-level message passing, making it well-suited for point-cloud observations and scalable simulation. This architecture can also be adapted to learn soft-body dynamics by replacing the rigid-body module (differentiable Kabsch alignment). 🎬See our video for more details. Many thanks to my amazing collaborators: Minghao Guo @GuoMh14, Haixu Wu @Haixu_Wu_1998, Doug Roble, Tuur Stuyck @TuurStuyck, and Wojciech Matusik @wojmatusik. Project page: people.csail.mit.edu/frankzydou/pro… Paper: people.csail.mit.edu/frankzydou/pro…
🥳 We are presenting Uncertainty-Aware Robotic World Model at the Workshop on World Models @iclr_conf in Rio 🇧🇷! sites.google.com/view/iclr-2026… We learn an RL policy from a neural network model with pure offline imagination from scratch and deploy on hardware! sites.google.com/view/uncertain…
Genesis AI has exited stealth for real with the announcement of GENE-26.5, a foundation model designed to achieve human-level physical manipulation. The system uses a proprietary dexterous hand and a tactile-sensing glove that is reportedly 100 times cheaper and 5 times more efficient at data collection than legacy teleoperation methods. Watch the system autonomously master tasks ranging from 20-step meal prep and wire harnessing to precision lab experiments and piano performance at 1x speed. 📽️🤖
I'm uploading a recording here to make up for the cut-off during our onsite presentation @NeurIPSConf #NeurIPS2025. 🙏We are deeply grateful for your previous support and the encouragement to upload this recording. ⭐️Despite the interruption, our work was recognized with an Outstanding Paper award.
Meta Platforms Inc. has acquired Assured Robot Intelligence, a startup developing artificial intelligence models for robots, as part of a major initiative to build humanoid technology. bloomberg.com/news/articles/…
Meta Platforms Inc. has acquired Assured Robot Intelligence, a startup developing artificial intelligence models for robots, as part of a major initiative to build humanoid technology. bloomberg.com/news/articles/…
