Adaptive Inference-Time Scaling via Cyclic Diffusion Search
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914115764617216 |
|---|---|
| author | Lee, Gyubin Bao, Truong Nhat Nguyen Yoon, Jaesik Lee, Dongwoo Kim, Minsu Bengio, Yoshua Ahn, Sungjin |
| author_facet | Lee, Gyubin Bao, Truong Nhat Nguyen Yoon, Jaesik Lee, Dongwoo Kim, Minsu Bengio, Yoshua Ahn, Sungjin |
| contents | Diffusion models have demonstrated strong generative capabilities across domains ranging from image synthesis to complex reasoning tasks. However, most inference-time scaling methods rely on fixed denoising schedules, limiting their ability to allocate computation based on instance difficulty or task-specific demands adaptively. We introduce the challenge of adaptive inference-time scaling-dynamically adjusting computational effort during inference-and propose Adaptive Bi-directional Cyclic Diffusion (ABCD), a flexible, search-based inference framework. ABCD refines outputs through bi-directional diffusion cycles while adaptively controlling exploration depth and termination. It comprises three components: Cyclic Diffusion Search, Automatic Exploration-Exploitation Balancing, and Adaptive Thinking Time. Experiments show that ABCD improves performance across diverse tasks while maintaining computational efficiency. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_14036 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Adaptive Inference-Time Scaling via Cyclic Diffusion Search Lee, Gyubin Bao, Truong Nhat Nguyen Yoon, Jaesik Lee, Dongwoo Kim, Minsu Bengio, Yoshua Ahn, Sungjin Machine Learning Artificial Intelligence Diffusion models have demonstrated strong generative capabilities across domains ranging from image synthesis to complex reasoning tasks. However, most inference-time scaling methods rely on fixed denoising schedules, limiting their ability to allocate computation based on instance difficulty or task-specific demands adaptively. We introduce the challenge of adaptive inference-time scaling-dynamically adjusting computational effort during inference-and propose Adaptive Bi-directional Cyclic Diffusion (ABCD), a flexible, search-based inference framework. ABCD refines outputs through bi-directional diffusion cycles while adaptively controlling exploration depth and termination. It comprises three components: Cyclic Diffusion Search, Automatic Exploration-Exploitation Balancing, and Adaptive Thinking Time. Experiments show that ABCD improves performance across diverse tasks while maintaining computational efficiency. |
| title | Adaptive Inference-Time Scaling via Cyclic Diffusion Search |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2505.14036 |