Search papers, labs, and topics across Lattice.
1
0
3
0
Skip task-specific reward engineering: Dex4D learns a single, generalizable dexterous manipulation policy in simulation that transfers zero-shot to real-world tasks by tracking object-centric points.
