Skywork UniPic 2.0: Building Kontext Model with Online RL for Unified Multimodal Model
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_ | 1866909997306216448 |
|---|---|
| author | Wei, Hongyang Xu, Baixin Liu, Hongbo Wu, Size Liu, Jie Peng, Yi Wang, Peiyu Liu, Zexiang He, Jingwen Xietian, Yidan Tang, Chuanxin Wang, Zidong Wei, Yichen Hu, Liang Jiang, Boyi Li, Wei He, Ying Liu, Yang Song, Xuchen Li, Yangguang Zhou, Yahui |
| author_facet | Wei, Hongyang Xu, Baixin Liu, Hongbo Wu, Size Liu, Jie Peng, Yi Wang, Peiyu Liu, Zexiang He, Jingwen Xietian, Yidan Tang, Chuanxin Wang, Zidong Wei, Yichen Hu, Liang Jiang, Boyi Li, Wei He, Ying Liu, Yang Song, Xuchen Li, Yangguang Zhou, Yahui |
| contents | Recent advances in multimodal models have demonstrated impressive capabilities in unified image generation and editing. However, many prominent open-source models prioritize scaling model parameters over optimizing training strategies, limiting their efficiency and performance. In this work, we present UniPic2-SD3.5M-Kontext, a 2B-parameter DiT model based on SD3.5-Medium, which achieves state-of-the-art image generation and editing while extending seamlessly into a unified multimodal framework. Our approach begins with architectural modifications to SD3.5-Medium and large-scale pre-training on high-quality data, enabling joint text-to-image generation and editing capabilities. To enhance instruction following and editing consistency, we propose a novel Progressive Dual-Task Reinforcement strategy (PDTR), which effectively strengthens both tasks in a staged manner. We empirically validate that the reinforcement phases for different tasks are mutually beneficial and do not induce negative interference. After pre-training and reinforcement strategies, UniPic2-SD3.5M-Kontext demonstrates stronger image generation and editing capabilities than models with significantly larger generation parameters-including BAGEL (7B) and Flux-Kontext (12B). Furthermore, following the MetaQuery, we connect the UniPic2-SD3.5M-Kontext and Qwen2.5-VL-7B via a connector and perform joint training to launch a unified multimodal model UniPic2-Metaquery. UniPic2-Metaquery integrates understanding, generation, and editing, achieving top-tier performance across diverse tasks with a simple and scalable training paradigm. This consistently validates the effectiveness and generalizability of our proposed training paradigm, which we formalize as Skywork UniPic 2.0. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_04548 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Skywork UniPic 2.0: Building Kontext Model with Online RL for Unified Multimodal Model Wei, Hongyang Xu, Baixin Liu, Hongbo Wu, Size Liu, Jie Peng, Yi Wang, Peiyu Liu, Zexiang He, Jingwen Xietian, Yidan Tang, Chuanxin Wang, Zidong Wei, Yichen Hu, Liang Jiang, Boyi Li, Wei He, Ying Liu, Yang Song, Xuchen Li, Yangguang Zhou, Yahui Computer Vision and Pattern Recognition Recent advances in multimodal models have demonstrated impressive capabilities in unified image generation and editing. However, many prominent open-source models prioritize scaling model parameters over optimizing training strategies, limiting their efficiency and performance. In this work, we present UniPic2-SD3.5M-Kontext, a 2B-parameter DiT model based on SD3.5-Medium, which achieves state-of-the-art image generation and editing while extending seamlessly into a unified multimodal framework. Our approach begins with architectural modifications to SD3.5-Medium and large-scale pre-training on high-quality data, enabling joint text-to-image generation and editing capabilities. To enhance instruction following and editing consistency, we propose a novel Progressive Dual-Task Reinforcement strategy (PDTR), which effectively strengthens both tasks in a staged manner. We empirically validate that the reinforcement phases for different tasks are mutually beneficial and do not induce negative interference. After pre-training and reinforcement strategies, UniPic2-SD3.5M-Kontext demonstrates stronger image generation and editing capabilities than models with significantly larger generation parameters-including BAGEL (7B) and Flux-Kontext (12B). Furthermore, following the MetaQuery, we connect the UniPic2-SD3.5M-Kontext and Qwen2.5-VL-7B via a connector and perform joint training to launch a unified multimodal model UniPic2-Metaquery. UniPic2-Metaquery integrates understanding, generation, and editing, achieving top-tier performance across diverse tasks with a simple and scalable training paradigm. This consistently validates the effectiveness and generalizability of our proposed training paradigm, which we formalize as Skywork UniPic 2.0. |
| title | Skywork UniPic 2.0: Building Kontext Model with Online RL for Unified Multimodal Model |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2509.04548 |