Architecture Decoupling Is Not All You Need For Unified Multimodal Model
Research paper by Dian Zheng and collaborators. Published 2025-11-27. Topics: multimodal models, image generation, understanding, task-specific multimodal interaction, cross-modal attention.
Resource information
- Authors
- Dian Zheng, Manyuan Zhang, Hongyu Li, Kai Zou, Hongbo Liu, Ziyu Guo, Kaituo Feng, Yexin Liu, Ying Luo, Yan Feng, Peng Pei, Xunliang Cai, Hongsheng Li
- Research topics
- multimodal models, image generation, understanding, task-specific multimodal interaction, cross-modal attention, Attention Interaction Alignment (AIA) loss, SFT, post-training stage, Qwen-VL, HunyuanImage, Emu3, Janus-Pro
- Publication date
- 2025-11-27
- Publication
- arXiv
- Last record update
- Nov 3, 2026