Diffusion Domain Expansion: Learning to Coordinate Pre-trained Diffusion Models
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918518100852736 |
|---|---|
| author | Lifar, Egor Savkin, Semyon Garipov, Timur Tong, Shangyuan Jaakkola, Tommi |
| author_facet | Lifar, Egor Savkin, Semyon Garipov, Timur Tong, Shangyuan Jaakkola, Tommi |
| contents | In this paper, we propose Diffusion Domain Expansion (DDE), a method that efficiently extends pre-trained diffusion models to generate larger objects and handle more complex conditioning beyond their original capabilities. Our method employs a compact trainable network designed to coordinate the denoised outputs of pre-trained diffusion models. We demonstrate that the coordinator can be universally simple while being capable of generalizing to domains larger than those observed during its training time. We evaluate DDE on long audio track generation and conditional image generation, demonstrating its applicability across domains. DDE outperforms other approaches to coordinated generation with diffusion models in qualitative and quantitative evaluations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_23275 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Diffusion Domain Expansion: Learning to Coordinate Pre-trained Diffusion Models Lifar, Egor Savkin, Semyon Garipov, Timur Tong, Shangyuan Jaakkola, Tommi Machine Learning In this paper, we propose Diffusion Domain Expansion (DDE), a method that efficiently extends pre-trained diffusion models to generate larger objects and handle more complex conditioning beyond their original capabilities. Our method employs a compact trainable network designed to coordinate the denoised outputs of pre-trained diffusion models. We demonstrate that the coordinator can be universally simple while being capable of generalizing to domains larger than those observed during its training time. We evaluate DDE on long audio track generation and conditional image generation, demonstrating its applicability across domains. DDE outperforms other approaches to coordinated generation with diffusion models in qualitative and quantitative evaluations. |
| title | Diffusion Domain Expansion: Learning to Coordinate Pre-trained Diffusion Models |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2605.23275 |