Search papers, labs, and topics across Lattice.
Affiliation:
6
0
6
27
TACL boosts legged locomotion success rates by over 56% on unstructured terrain by dynamically adapting training tasks to the evolving capabilities of the policy.
TeleDexter achieves a remarkable 75% success rate in dexterous teleoperation tasks, where existing systems fail, showcasing a leap towards human-level control in robotic manipulation.
Unlock the potential of readily available internet videos to train 3D scene understanding models, achieving strong zero-shot performance and paving the way for more capable systems.
A new humanoid teleoperation system, OmniClone, achieves state-of-the-art control fidelity and generalization while drastically reducing computational cost, thanks to a diagnostic benchmark that exposes failure modes overlooked by aggregate metrics.
Current VLM benchmarks don't adequately test embodied spatial reasoning, so ESPIRE fills the gap with a physically grounded environment and generative task formulation.
Forget scaling laws: reinforcement fine-tuning with verifiable rewards lets a 4B parameter model beat an 8B parameter model on challenging 3D scene understanding tasks.
