arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~

高校专区

Beihang University(北京航空航天大学)

2026-03-27 至 2026-03-27 共收录 4
2603.25170 2026-03-27 cs.CV cs.AI

Knowledge-Guided Adversarial Training for Infrared Object Detection via Thermal Radiation Modeling

基于热辐射建模的知识引导对抗训练用于红外目标检测

Shiji Zhao, Shukun Xiong, Maoxun Yuan, Yao Huang, Ranjie Duan, Qing Guo, Jiansheng Chen, Haibin Duan, Xingxing Wei

机构 * Institute of Artificial Intelligence, Beihang University(北京航空航天大学人工智能研究院) Security Department, Alibaba Group(阿里巴巴集团安全部) School of Computer Science, Nankai University(南开大学计算机学院) School of Computer and Communication Engineering, University of Science and Technology Beijing(北京科技大学计算机与通信工程学院) School of Automation Science and Electrical Engineering, Beihang University(北京航空航天大学自动化科学与电气工程学院)

AI总结 本文提出KGAT方法,利用红外物理知识提升目标检测的鲁棒性,通过热辐射关系建模优化对抗训练,实验表明在三个数据集上提升了准确性和抗攻击能力。

Comments Accepted for publication in the International Journal of Computer Vision (IJCV)

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.02787 2026-03-27 cs.RO cs.CV

Diagnose, Correct, and Learn from Manipulation Failures via Visual Symbols

通过视觉符号诊断、纠正并学习操纵失败

Xianchao Zeng, Xinyu Zhou, Youcheng Li, Jiayou Shi, Tianle Li, Liangming Chen, Lei Ren, Yong-Lu Li

机构 * Beihang University(北京航空航天大学) Shanghai Innovation Institute(上海创新研究院) Southern University of Science and Technology(南方科技大学) Shanghai Jiao Tong University(上海交通大学)

AI总结 本文提出ViFailback框架,通过视觉符号提升标注效率,并释放大规模数据集和基准测试,验证了框架在故障诊断和纠正中的有效性。

Comments Accepted by CVPR 2026. Project Website: https://x1nyuzhou.github.io/vifailback.github.io/

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.25083 2026-03-27 cs.CV cs.AI

Learning domain-invariant features through channel-level sparsification for Out-Of Distribution Generalization

通过通道级稀疏化学习领域不变特征以实现分布外泛化

Haoran Pei, Yuguang Yang, Kexin Liu, Juan Zhang, Baochang Zhang

机构 * School of Automation Science and Electrical Engineering, Beihang University(北京航空航天大学自动化科学与电气工程学院) School of Artificial Intelligence, Beihang University(北京航空航天大学人工智能学院) School of Electronic Information Engineering, Beihang University(北京航空航天大学电子信息工程学院)

AI总结 本文提出Hierarchical Causal Dropout方法,通过通道级因果掩码实现特征稀疏化,以分离因果特征与虚假特征,提升分布外泛化能力。

详情

展开后加载摘要…

URL PDF HTML 收藏
2410.10700 2026-03-27 cs.CL cs.AI

LLMs know their vulnerabilities: Uncover Safety Gaps through Natural Distribution Shifts

大语言模型知其弱点:通过自然分布偏移揭示安全漏洞

Qibing Ren, Hao Li, Dongrui Liu, Zhanxu Xie, Xiaoya Lu, Yu Qiao, Lei Sha, Junchi Yan, Lizhuang Ma, Jing Shao

机构 * MoE Key Lab of Artificial Intelligence, Shanghai Jiao Tong University(人工智能大模型关键实验室,上海交通大学) Beihang University(北京航空航天大学) Shanghai Artificial Intelligence Laboratory(上海人工智能实验室)

AI总结 本文研究大语言模型对自然分布偏移的脆弱性,提出ActorBreaker攻击方法,通过识别有毒提示相关角色构建多轮提示,揭示模型安全漏洞,并构建安全数据集提升鲁棒性。

Comments ACL 2025 main conference. Code is available at https://github.com/AI45Lab/ActorAttack

详情

展开后加载摘要…

URL PDF HTML 收藏