SIGMA: Selective-Interleaved Generation with Multi-Attribute Tokens
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914314058727424 |
|---|---|
| author | Zhang, Xiaoyan Bai, Zechen Wang, Haofan Song, Yiren |
| author_facet | Zhang, Xiaoyan Bai, Zechen Wang, Haofan Song, Yiren |
| contents | Recent unified models such as Bagel demonstrate that paired image-edit data can effectively align multiple visual tasks within a single diffusion transformer. However, these models remain limited to single-condition inputs and lack the flexibility needed to synthesize results from multiple heterogeneous sources. We present SIGMA (Selective-Interleaved Generation with Multi-Attribute Tokens), a unified post-training framework that enables interleaved multi-condition generation within diffusion transformers. SIGMA introduces selective multi-attribute tokens, including style, content, subject, and identity tokens, which allow the model to interpret and compose multiple visual conditions in an interleaved text-image sequence. Through post-training on the Bagel unified backbone with 700K interleaved examples, SIGMA supports compositional editing, selective attribute transfer, and fine-grained multimodal alignment. Extensive experiments show that SIGMA improves controllability, cross-condition consistency, and visual quality across diverse editing and generation tasks, with substantial gains over Bagel on compositional tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_07564 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SIGMA: Selective-Interleaved Generation with Multi-Attribute Tokens Zhang, Xiaoyan Bai, Zechen Wang, Haofan Song, Yiren Computer Vision and Pattern Recognition Recent unified models such as Bagel demonstrate that paired image-edit data can effectively align multiple visual tasks within a single diffusion transformer. However, these models remain limited to single-condition inputs and lack the flexibility needed to synthesize results from multiple heterogeneous sources. We present SIGMA (Selective-Interleaved Generation with Multi-Attribute Tokens), a unified post-training framework that enables interleaved multi-condition generation within diffusion transformers. SIGMA introduces selective multi-attribute tokens, including style, content, subject, and identity tokens, which allow the model to interpret and compose multiple visual conditions in an interleaved text-image sequence. Through post-training on the Bagel unified backbone with 700K interleaved examples, SIGMA supports compositional editing, selective attribute transfer, and fine-grained multimodal alignment. Extensive experiments show that SIGMA improves controllability, cross-condition consistency, and visual quality across diverse editing and generation tasks, with substantial gains over Bagel on compositional tasks. |
| title | SIGMA: Selective-Interleaved Generation with Multi-Attribute Tokens |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.07564 |