arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

Huazhong University of Science and Technology(华中科技大学)

2026-01-29 至 2026-01-29 共收录 3
2601.20325 2026-01-29 cs.CR cs.CV

UnlearnShield: Shielding Forgotten Privacy against Unlearning Inversion

UnlearnShield: 保护被遗忘隐私 against 反向学习

Lulu Xue, Shengshan Hu, Wei Lu, Ziqi Zhou, Yufei Song, Jianhong Cheng, Minghui Li, Yanjun Zhang, Leo Yu Zhang

机构 * Huazhong University of Science and Technology(华中科技大学) Institute of Guizhou Aerospace Measuring and Testing Technology(贵州航天测量测试技术研究院) University of Technology Sydney(悉尼科技大学) Griffith University(格里菲斯大学)

AI总结 UnlearnShield通过引入方向扰动和约束模块,有效降低反向学习反向的风险,同时保持模型准确性和实用性。

Comments This work has been accepted by ICASSP 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.20707 2026-01-29 cs.CV

Mixing Importance with Diversity: Joint Optimization for KV Cache Compression in Large Vision-Language Models

融合重要性与多样性:KV缓存压缩的联合优化

Xuyang Liu, Xiyan Gui, Yuchao Zhang, Linfeng Zhang

机构 * EPIC Lab, Shanghai Jiao Tong University(上海交通大学) Sichuan University(四川大学) Huazhong University of Science and Technology(华中科技大学)

AI总结 MixKV通过融合重要性与多样性,优化KV缓存压缩,提升多模态模型的存储效率与推理性能。

Comments Accepted by ICLR 2026. Our code is available at https://github.com/xuyang-liu16/MixKV

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.12605 2026-01-29 cs.CV

WaterFlow: Explicit Physics-Prior Rectified Flow for Underwater Saliency Mask Generation

WaterFlow: 基于显式物理先验的 rectified 流用于水下显著性掩码生成

Runting Li, Shijie Lian, Hua Li, Yutong Li, Wenhui Wu, Sam Kwong

机构 * Hainan University, China(海南大学) Huazhong University of Science and Technology, China(华中科技大学) Shenzhen University, China(深圳大学) Lingnan University, Hong Kong(岭南大学)

AI总结 WaterFlow通过引入显式物理先验和时间维度建模,提升水下显著性目标检测的性能,在USOD10K数据集上取得显著成效。

详情

展开后加载摘要…

URL PDF HTML 收藏