6 papers
Debate2Create: Robot Co-design via Multi-Agent LLM Debate
Kevin Qiu, Marek Cygan
We introduce Debate2Create (D2C), a multi-agent LLM framework that formulates robot co-design as structured, iterative debate grounded in physics-based evaluation. A design agent a…
Kine2Go: Kinematic dataset for the Unitree Go2 robot with diverse gaits and motions
WÅadysÅaw PaÅucki, PaweÅ Siwak, Krzysztof Ciebiera +1
The recent popularity of robotics, combined with the steadily decreasing cost of robotic hardware, has lowered the entry barrier to robotics research and enabled rapid advancements…
Reward-Conditioned Reinforcement Learning
Michal Nauman, Marek Cygan, Pieter Abbeel
Single-task RL agents are typically trained under a fixed reward function, which limits their robustness to reward misspecification and their ability to adapt to changing preferenc…
Swim2Real: VLM-Guided System Identification for Sim-to-Real Transfer
Kevin Qiu, Kyle Walker, Mike Y. Michelis +2
We present Swim2Real, a pipeline that calibrates a 16-parameter robotic fish simulator from swimming videos using vision-language model (VLM) feedback, requiring no hand-designed s…
Vid2Sid: Videos Can Help Close the Sim2Real Gap
Kevin Qiu, Yu Zhang, Marek Cygan +1
Calibrating a robot simulator's physics parameters (friction, damping, material stiffness) to match real hardware is often done by hand or with black-box optimizers that reduce err…
Bigger, Regularized, Categorical: High-Capacity Value Functions are Efficient Multi-Task Learners
Michal Nauman, Marek Cygan, Carmelo Sferrazza +2
Recent advances in language modeling and vision stem from training large models on diverse, multi-task data. This paradigm has had limited impact in value-based reinforcement learn…