1 paper
Yeyao Ma, Chen Li, Xiaosong Zhang +2
Post-training of flow matching models-aligning the output distribution with a high-quality target-is mathematically equivalent to imitation learning. While Supervised Fine-Tuning m…