Search papers, labs, and topics across Lattice.
3
0
4
High-quality dense rewards can transform robotic manipulation, propelling task success rates from 50% to near perfection in real-world applications.
Uncovering that CoT reasoning transitions from uncertainty to high reliability opens the door to more efficient inference strategies that can save computational resources without sacrificing accuracy.
Training VLA policies without human demos is now feasible, with LEGS achieving better performance than traditional methods at a fraction of the cost.
