1 paper · 1 filter
Ruoyu Wang, Boye Niu, Xiangxin Zhou +3
Aligning text-to-image flow matching models with human preferences via direct reward backpropagation is sample-efficient but hampered by two well-known pathologies: activations can…