Search papers, labs, and topics across Lattice.
1
0
3
3
VLA models can now achieve significantly higher performance and stability in embodied AI tasks by combining a Mixture-of-Transformers architecture with a novel flow-matching-based reinforcement learning algorithm.
