Search papers, labs, and topics across Lattice.
Fudan University
1
0
2
Tree search on on-policy trajectories dramatically boosts policy gradient learning by uncovering rare high-return paths, outperforming PPO on Atari and MuJoCo with minimal bias increase.
