Search papers, labs, and topics across Lattice.
1
0
3
Flow-based imitation learning can be significantly improved by distilling both rewards and actions on-policy, enabling more robust and generalizable policies, especially with limited or noisy demonstrations.
