This work formalizes skill optimization via Zeroth-Order (ZO) optimization, mapping classical counterparts (central difference, trust regions) to recent literature to establish three principles for convergence and generalization: file-system-based trajectory exploration, consensus attribute mining, and independent vali...
Yifei Shen, Bo Li, Xinjie Zhang· arXiv.org· 10 citations
Mage-VL is presented, an efficient codec-native streaming foundation model for real-time multimodal understanding and interaction and establishes AI4AI data pipelines encompassing prompt-code joint optimization for multimodal captioning and AI-driven performance diagnosis to guide training recipes.
Results show that a fixed-weight, self-evolving harness can revise, recover, and accumulate verified approaches while producing structured trajectories for future supervised and reinforcement learning.
Boxiu Li, Zi-Mo Wen, Yi-Jia Fan et al.· 2 citations· ⚡1
BiVidGen, a hybrid framework where an MLLM first generates semantic visual tokens and a DiT renders videos conditioned on both text and these tokens via multi-layer cross-attention is proposed.
Yan-Bo Ding, Yijia Fan, Caihua Shan et al.· 0 citations
The results show that careful tokenizer--backbone--system co-design can deliver strong high-resolution generation and editing within an efficient 4B model family.
