Search papers, labs, and topics across Lattice.
1
0
2
ELBO-based reinforcement learning, previously dismissed for visual generation, can actually outperform MDP-based methods for aligning denoising generative models with human preferences.
