Search papers, labs, and topics across Lattice.
2
0
6
0
It is shown that the minimum aggregation is essential: mean aggregation can be exploited, saturating the proxy reward while degrading accuracy, and the minimum aggregation is essential for PROSE, Process Reward Guided Self-Training.
Autoregressive image and video generators can get a surprising boost in quality *without* retraining, just by steering them towards outputs that better support their own past creations.
