← 研踪 PaperTrailCVF · CVPR 2026:cvf:c4d9415525c2efc69f33613e67aadd03 · 记录 #1
Mixture of States: Routing Token-Level Dynamics for Multimodal Generation
Liu, HaozheLiu, DingZhuge, MingchenZhou, ZijianXie, TianHe, SenYang, YukangLiu, ShumingCong, YurenGuo, JiadongXu, HongyuXu, KeNg, Kam-WohPerez, Juan C.Perez-Rua, Juan-ManuelXiang, TaoLiu, WeiLiu, ShikunSchmidhuber, Jürgen 来源日期（仅到年份） 2026
来源记录日期 2026
雷达观测 2026-11-05 02:27
摘要
We introduce MoS (Mixture of States), a novel fusion paradigm for multimodal diffusion models that merges modalities using flexible, state-based interactions. The core of MoS is a learnable, token-wise router that creates denoising timestep- and input-dependent interactions between modalities' hidden states, precisely aligning token-level features with the diffusion trajectory. This router sparsely selects the top-k hidden states and is trained with an epsilon-greedy strategy, efficiently selecting contextual features with minimal learnable parameters and negligible computational overhead. We validate our design with text-to-image generation (MoS-Image) and editing (MoS-Editing), which achieve state-of-the-art results. With only 3B to 5B parameters, our models match or surpass counterparts up to 4xlarger. These findings establish MoS as a flexible and compute-efficient paradigm for scaling multimodal diffusion models. The project homepage is available at https://haozheliu-st.github.io/mos-homepage/
论文发现出处
这些页面提供论文线索，不代表同行评审或质量认证。
观测到的版本
平台记录号代表本站观测到的元数据变化，不是作者修订号。来源日期仅到月份或年份时，按该月或该年首日存档。
记录 #1 · 2026 · 最新观测版本
Mixture of States: Routing Token-Level Dynamics for Multimodal Generation
We introduce MoS (Mixture of States), a novel fusion paradigm for multimodal diffusion models that merges modalities using flexible, state-based interactions. The core of MoS is a learnable, token-wise router that creates denoising timestep- and input-dependent interactions between modalities' hidden states, precisely aligning token-level features with the diffusion trajectory. This router sparsely selects the top-k hidden states and is trained with an epsilon-greedy strategy, efficiently selecting contextual features with minimal learnable parameters and negligible computational overhead. We validate our design with text-to-image generation (MoS-Image) and editing (MoS-Editing), which achieve state-of-the-art results. With only 3B to 5B parameters, our models match or surpass counterparts up to 4xlarger. These findings establish MoS as a flexible and compute-efficient paradigm for scaling multimodal diffusion models. The project homepage is available at https://haozheliu-st.github.io/mos-homepage/
查看此版本解读 →查看此版本原文 ↗