Inference-time Scaling of Diffusion Models through Classical Search
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866916989446914048 |
|---|---|
| author | Zhang, Xiangcheng Lin, Haowei Ye, Haotian Zou, James Ma, Jianzhu Liang, Yitao Du, Yilun |
| author_facet | Zhang, Xiangcheng Lin, Haowei Ye, Haotian Zou, James Ma, Jianzhu Liang, Yitao Du, Yilun |
| contents | Classical search algorithms have long underpinned modern artificial intelligence. In this work, we tackle the challenge of inference-time control in diffusion models -- adapting generated outputs to meet diverse test-time objectives -- using principles from classical search. We propose a general framework that orchestrates local and global search to efficiently navigate the generative space. It employs a theoretically grounded local search via annealed Langevin MCMC and performs compute-efficient global exploration using breadth-first and depth-first tree search. We evaluate our approach on a range of challenging domains, including planning, offline reinforcement learning, and image generation. Across all tasks, we observe significant gains in both performance and efficiency. These results show that classical search provides a principled and practical foundation for inference-time scaling in diffusion models. Project page at https://diffusion-inference-scaling.github.io/. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_23614 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Inference-time Scaling of Diffusion Models through Classical Search Zhang, Xiangcheng Lin, Haowei Ye, Haotian Zou, James Ma, Jianzhu Liang, Yitao Du, Yilun Machine Learning Classical search algorithms have long underpinned modern artificial intelligence. In this work, we tackle the challenge of inference-time control in diffusion models -- adapting generated outputs to meet diverse test-time objectives -- using principles from classical search. We propose a general framework that orchestrates local and global search to efficiently navigate the generative space. It employs a theoretically grounded local search via annealed Langevin MCMC and performs compute-efficient global exploration using breadth-first and depth-first tree search. We evaluate our approach on a range of challenging domains, including planning, offline reinforcement learning, and image generation. Across all tasks, we observe significant gains in both performance and efficiency. These results show that classical search provides a principled and practical foundation for inference-time scaling in diffusion models. Project page at https://diffusion-inference-scaling.github.io/. |
| title | Inference-time Scaling of Diffusion Models through Classical Search |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2505.23614 |