Flash-BoN: Instant Drafts for Inference-Time Scaling in Diffusion Models
Flash-BoN improves text-to-image generation efficiency by using inexpensive draft candidates generated through timestep truncation, layer skipping, and activation proxies, followed…
Hugging Face · Daily Papers
·Ruchit Rawal, Reza Shirkavand
·
·▲ 4 upvotes
Este artigo está em destaque na seleção diária de papers do Hugging Face, curada pela comunidade de pesquisa em IA.
Autores: Ruchit Rawal, Reza Shirkavand, Sayak Paul, Yuxin Wen, Heng Huang, Yizheng Chen
- 4 upvotes da comunidade
- Temas: Best-of-N, guided search methods, denoising steps, function evaluations, wall-clock evaluation, Flash-BoN
Resumo
Resumo original (em inglês), extraído do paper:
Flash-BoN improves text-to-image generation efficiency by using inexpensive draft candidates generated through timestep truncation, layer skipping, and activation proxies, followed by multi-stage verification that outperforms existing methods under fixed wall-clock budgets.Onde ler
// relacionados
Leia também
Editorial
EditBridge: edição de imagem em 4K sem inventar o que não estava lá
Blog
Coarse-to-Fine Multi-Resolution Diffusion Models for Trajectory Generation in Urban Systems
Blog
DiSCO: Defending text-to-image generation through distribution-guided contrastive prompt optimization
Blog