Diagnosing and Correcting Concept Omission in Multimodal Diffusion Transformers
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_ | 1866914611312197632 |
|---|---|
| author | Baek, Kanghyun Lew, Jaihyun Shin, Chaehun Lee, Jungbeom Yoon, Sungroh |
| author_facet | Baek, Kanghyun Lew, Jaihyun Shin, Chaehun Lee, Jungbeom Yoon, Sungroh |
| contents | Multimodal Diffusion Transformers (MM-DiTs) have achieved remarkable progress in text-to-image generation, yet they frequently suffer from concept omission, where specified objects or attributes fail to emerge in the generated image. By performing linear probing on text tokens, we demonstrate that text embeddings can distinguish a characteristic `omission signal' representing the absence of target concepts. Leveraging this insight, we propose Omission Signal Intervention (OSI), which amplifies the omission signal to actively catalyze the generation of missing concepts. Comprehensive experiments on FLUX.1-Dev and SD3.5-Medium demonstrate that OSI significantly alleviates concept omission even in extreme scenarios. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_14270 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Diagnosing and Correcting Concept Omission in Multimodal Diffusion Transformers Baek, Kanghyun Lew, Jaihyun Shin, Chaehun Lee, Jungbeom Yoon, Sungroh Computer Vision and Pattern Recognition Multimodal Diffusion Transformers (MM-DiTs) have achieved remarkable progress in text-to-image generation, yet they frequently suffer from concept omission, where specified objects or attributes fail to emerge in the generated image. By performing linear probing on text tokens, we demonstrate that text embeddings can distinguish a characteristic `omission signal' representing the absence of target concepts. Leveraging this insight, we propose Omission Signal Intervention (OSI), which amplifies the omission signal to actively catalyze the generation of missing concepts. Comprehensive experiments on FLUX.1-Dev and SD3.5-Medium demonstrate that OSI significantly alleviates concept omission even in extreme scenarios. |
| title | Diagnosing and Correcting Concept Omission in Multimodal Diffusion Transformers |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2605.14270 |