Search papers, labs, and topics across Lattice.
Cornell University
1
0
3
Training early-stage rankers in retrieval systems doesn't have to be a high-variance nightmare: CA-PG offers a stable, faster-converging alternative to vanilla policy gradients.