Exploring the Design Space of Reward Backpropagation for Flow Matching

Exploring the Design Space of Reward Backpropagation for Flow Matching

FlowBP addresses limitations in flow matching model alignment by using a surrogate trajectory framework that reduces memory usage and gradient chaining while maintaining performanc…

Hugging Face · Daily Papers ·Ruoyu Wang, Boye Niu · ·▲ 6 upvotes

Este artigo está em destaque na seleção diária de papers do Hugging Face, curada pela comunidade de pesquisa em IA.

Autores: Ruoyu Wang, Boye Niu, Xiangxin Zhou, Yushi Huang, Tongliang Liu, Chi Zhang

  • 6 upvotes da comunidade
  • Temas: flow matching models, direct reward backpropagation, Jacobian products, backward trajectory, surrogate trajectory, cached rollout

Resumo

Resumo original (em inglês), extraído do paper:

FlowBP addresses limitations in flow matching model alignment by using a surrogate trajectory framework that reduces memory usage and gradient chaining while maintaining performance across multiple text-to-image models.

Ler o paper completo no Hugging Face →

compartilhar: