Real2Sim2Real presents a promising opportunity for zero-shot robot deployment 🤖, but how does this paradigm translate to tricky, contact-rich applications such as the game of billiards 🎱?
We argue that an agent can learn the underlying dynamics parameters through interaction!
Can an agent learn a pool table’s physics by playing on it? 🎱
While agents can close the visual gap, the difference between real world and sim dynamics remains significant. For CoRL late breaking🤠, we let an agent take real shots to try and close this real2sim dynamics gap: cue-dynamics.github.io
Very fun short project with @XilunZhangXZ @MatthewFoutter @Luis_A_Pabon @drmapavone
Matthew Foutter retweeted
✨ Test-Time Scaling for Robotics ✨
Excited to release 🤖 RoboMonkey, which characterizes test-time scaling laws for Vision-Language-Action (VLA) models and introduces a framework that significantly improves the generalization and robustness of VLAs!
🧵(1 / N)
🌐 Website: robomonkey-vla.github.io
💻 Code: github.com/robomonkey-vla/Ro…
🗂️ Datasets and Models: huggingface.co/robomonkey-vl…
🚀 Serving Engine: github.com/robomonkey-vla/sg…
📄 Paper: arxiv.org/abs/2506.17811
