Search papers, labs, and topics across Lattice.
1
0
3
Skip the costly human annotations: PromptEcho distills reward signals directly from frozen VLMs to boost text-to-image RL, achieving state-of-the-art results without any reward model training.
