Search papers, labs, and topics across Lattice.
1
0
3
9
Outcome-only RL can train highly effective process supervision models by rewarding accurate outcome prediction, even without explicit step-level feedback.
