ControlFlow3D is proposed, a framework that distills knowledge from pretrained foundation models into a lightweight generation network during training, enabling strong performance without multi-view images available at test time, and performs conditional flow matching in a structured latent space defined by a frozen 3D...
Yuang Liu, Zhi Zuo, Zheng-Kai Zhao et al.· 0 citations
MultiCompose is presented, a composition framework that decouples per-concept personalization from multi-subject inference and MSP-Bench is introduced, a benchmark that jointly evaluates identity fidelity, attribute binding accuracy, and attribute misalignment through a dual-pathway protocol.
GrainGS is a dynamic Gaussian framework that combines a hierarchical anchor scaffold with per-Gaussian deformation that achieves high reconstruction quality, real-time novel view synthesis, and compact storage.