Search papers, labs, and topics across Lattice.
1
0
3
1
Ditch the costly preference pairs and reward modeling: FAIL leverages adversarial imitation learning to fine-tune flow matching models, achieving state-of-the-art image generation with minimal data.
