Search papers, labs, and topics across Lattice.
1
0
2
Themis can train RL models that align better with true rewards than existing methods, all while being user-friendly and scalable for large experiments.