Search papers, labs, and topics across Lattice.
1
0
3
A 7B model, trained with reinforcement learning to reason about robotic manipulation, outperforms 72B-scale general MLLMs and even OpenAI's closed-source models in process supervision and failure detection.
