arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

University of Chinese Academy of Sciences(中国科学院大学)

2025-12-29 至 2025-12-29 共收录 5
2512.21857 2025-12-29 cs.CV

Fast Inference of Visual Autoregressive Model with Adjacency-Adaptive Dynamical Draft Trees

视觉自回归模型快速推断的邻接自适应动态草案树

Haodong Lei, Hongsong Wang, Xin Geng, Liang Wang, Pan Zhou

机构 * School of Computer Science and Engineering, Southeast University(东南大学计算机科学与工程学院) Key Laboratory of New Generation Artificial Intelligence Technology and Its Interdisciplinary Applications (Southeast University), Ministry of Education, China(东南大学新一代人工智能技术及交叉应用重点实验室(教育部)) New Laboratory of Pattern Recognition (NLPR), State Key Laboratory of Multimodal Artificial Intelligence Systems (MAIS), Institute of Automation, Chinese Academy of Sciences (CASIA)(模式识别新实验室(NLPR),多模态人工智能系统国家重点实验室(MAIS),中国科学院自动化研究所(CASIA)) School of Artificial Intelligence, University of Chinese Academy of Sciences(中国科学院大学人工智能学院) School of Computing and Information Systems, Singapore Management University(新加坡国立大学计算机与信息学院)

AI总结 ADT-Tree通过邻接自适应动态草案树提升视觉自回归模型的推断速度,实现3.13倍和3.05倍的加速。

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.21683 2025-12-29 cs.CV

Contrastive Graph Modeling for Cross-Domain Few-Shot Medical Image Segmentation

对比图模型用于跨域少样本医学图像分割

Yuntian Bo, Tao Zhou, Zechao Li, Haofeng Zhang, Ling Shao

机构 * School of Computer Science and Engineering, Nanjing University of Science and Technology(计算机科学与工程学院,南京理工大学) University of Chinese Academy of Sciences(中国科学院大学)

AI总结 本文提出对比图模型,通过结构一致性先验和子图匹配解码机制,在跨域少样本医学图像分割中实现高精度与强源域准确性。

Comments Accepted to IEEE Transactions on Medical Imaging (T-MI), 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.21586 2025-12-29 cs.LG

Videos are Sample-Efficient Supervisions: Behavior Cloning from Videos via Latent Representations

视频是高效的监督:通过潜在表示从视频中进行行为克隆

Xin Liu, Haoran Li, Dongbin Zhao

机构 * State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences(多模态人工智能系统国家重点实验室,自动化研究所,中国科学院) School of Artificial Intelligence, University of Chinese Academy of Sciences(人工智能学院,中国科学院大学)

AI总结 本文提出通过潜在表示从视频中进行行为克隆的方法,实现了高效样本的视觉策略学习,无需其他专家监督。

Journal ref NeurIPS 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.25244 2025-12-29 cs.LG

BSFA: Leveraging the Subspace Dichotomy to Accelerate Neural Network Training

BSFA:利用子空间二元性加速神经网络训练

Wenjie Zhou, Bohan Wang, Wei Chen, Xueqi Cheng

机构 * State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences(人工智能安全国家重点实验室,计算技术研究所,中国科学院) University of Chinese Academy of Sciences(中国科学院大学) Alibaba Group(阿里巴巴集团)

AI总结 BSFA通过子空间二元性加速神经网络训练,利用PCA估计子空间并分块策略提升收敛速度和稳定性。

Comments 16 pages

详情

展开后加载摘要…

URL PDF HTML 收藏
2507.06558 2025-12-29 cs.LG cs.AI

The Primacy of Magnitude in Low-Rank Adaptation

低秩适应中的主导性

Zicheng Zhang, Haoran Li, Yifeng Zhang, Guoqiang Gong, Jiaxing Wang, Junxing Hu, Pengzhang Liu, Qixia Jiang

机构 * University of Chinese Academy of Sciences(中国科学院大学)

AI总结 本文提出LoRAM,一种基于幅度驱动的低秩适应初始化方法,通过优化更新幅度提升效率并匹配或超越传统谱初始化方法。

Comments Accepted by NeurIPS 2025, spotlight

Journal ref NeurIPS 2025

详情

展开后加载摘要…

URL PDF HTML 收藏