DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918199063216128 |
|---|---|
| author | He, Xiaoxiao Dao, Quan Han, Ligong Wen, Song Bai, Minhao Liu, Di Zhang, Han Min, Martin Renqiang Juefei-Xu, Felix Tan, Chaowei Liu, Bo Li, Kang Li, Hongdong Huang, Junzhou Ahmed, Faez Srivastava, Akash Metaxas, Dimitris |
| author_facet | He, Xiaoxiao Dao, Quan Han, Ligong Wen, Song Bai, Minhao Liu, Di Zhang, Han Min, Martin Renqiang Juefei-Xu, Felix Tan, Chaowei Liu, Bo Li, Kang Li, Hongdong Huang, Junzhou Ahmed, Faez Srivastava, Akash Metaxas, Dimitris |
| contents | Discrete diffusion models have achieved success in tasks like image generation and masked language modeling but face limitations in controlled content editing. We introduce DICE (Discrete Inversion for Controllable Editing), the first approach to enable precise inversion for discrete diffusion models, including multinomial diffusion and masked generative models. By recording noise sequences and masking patterns during the reverse diffusion process, DICE enables accurate reconstruction and flexible editing of discrete data without the need for predefined masks or attention manipulation. We demonstrate the effectiveness of DICE across both image and text domains, evaluating it on models such as VQ-Diffusion, Paella, and RoBERTa. Our results show that DICE preserves high data fidelity while enhancing editing capabilities, offering new opportunities for fine-grained content manipulation in discrete spaces. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_08207 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models He, Xiaoxiao Dao, Quan Han, Ligong Wen, Song Bai, Minhao Liu, Di Zhang, Han Min, Martin Renqiang Juefei-Xu, Felix Tan, Chaowei Liu, Bo Li, Kang Li, Hongdong Huang, Junzhou Ahmed, Faez Srivastava, Akash Metaxas, Dimitris Computer Vision and Pattern Recognition Machine Learning Discrete diffusion models have achieved success in tasks like image generation and masked language modeling but face limitations in controlled content editing. We introduce DICE (Discrete Inversion for Controllable Editing), the first approach to enable precise inversion for discrete diffusion models, including multinomial diffusion and masked generative models. By recording noise sequences and masking patterns during the reverse diffusion process, DICE enables accurate reconstruction and flexible editing of discrete data without the need for predefined masks or attention manipulation. We demonstrate the effectiveness of DICE across both image and text domains, evaluating it on models such as VQ-Diffusion, Paella, and RoBERTa. Our results show that DICE preserves high data fidelity while enhancing editing capabilities, offering new opportunities for fine-grained content manipulation in discrete spaces. |
| title | DICE: Discrete Inversion Enabling Controllable Editing for Multinomial Diffusion and Masked Generative Models |
| topic | Computer Vision and Pattern Recognition Machine Learning |
| url | https://arxiv.org/abs/2410.08207 |