arXivDaily arXiv每日学术速递 周一至周五更新

期刊&会议

IEEE TPAMI

IEEE Transactions on Pattern Analysis and Machine Intelligence · 期刊 · Computer Vision

共收录 1623 篇
2611.15160 2026-11-13 cs.IR 新提交

Measuring, Revealing and Mitigating Harmful Content Propagation in Recommender Systems

测量、揭示与缓解推荐系统中的有害内容传播

Xiaokun Zhang, Youlin Wu, Junyu Lu, Jiamin Chen, Zhaochun Ren, Hongfei Lin, Fenglong Ma, Xue Liu, Chen Ma

机构 * School of Computer Science and Technology, Dalian University of Technology(大连理工大学计算机科学与技术学院) ; Department of Computer Science, MBZUAI(穆罕默德·本·扎耶德人工智能大学计算机系) ; School of Computer Science, McGill University(麦吉尔大学计算机学院)

AI总结 本研究针对推荐系统中有害内容被忽视的问题,构建标注数据集与评估指标,提出通用框架MitiHarm,可在保障用户满意度的同时缓解有害内容传播。

Comments This work has been acceptted by IEEE TPAMI

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2611.12975 2026-11-11 cs.CV cs.LG 新提交

Attribute-Evidence Learning with Multimodal Prompt for Generalized Category Discovery

面向广义类别发现的多模态提示属性-证据学习

Wei Zhang, Zhu Teng, Zhixiang Chi, Junnan Zou, Wenxin Luo, Linfeng Ye, Jianping Fan, Konstantions N. Plataniotis

机构 * Beijing Jiaotong University(北京交通大学) ; University of Toronto(多伦多大学) ; FuDan University(复旦大学)

AI总结 提出AptM-GCD多模态框架,通过属性证据学习和多模态提示学习解决广义类别发现中属性不可靠的问题,在六个数据集上达当前最优性能且泛化能力强。

Comments IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2604.00955 2026-11-10 cs.CV 版本更新

Enhancing Gradient Inversion Attacks in Federated Learning via Hierarchical Feature Optimization

通过分层特征优化增强联邦学习中的梯度反向攻击

Hao Fang, Wenbo Yu, Bin Chen, Xuan Wang, Shu-Tao Xia, Qing Liao, Ke Xu

机构 * Tsinghua Shenzhen International Graduate School, Tsinghua University(清华大学深圳国际研究生院) ; School of Computer Science and Technology, Harbin Institute of Technology(哈尔滨工业大学计算机科学与技术学院) ; Department of Computer Science and Technology, Tsinghua University(清华大学计算机科学与技术系)

AI总结 本文提出GIFD方法,通过分解GAN模型并搜索中间层的分层特征,提升梯度反向攻击的表达能力和泛化能力,同时引入正则化约束和OOD场景下的标签映射技术,实现像素级重建。

Comments TPAMI-2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2611.09754 2026-11-09 eess.IV 新提交

Tomographic Sparse View Selection using the View Covariance Loss

基于视图协方差损失的层析成像稀疏视图选择

Jingsong Lin, Amirkoushyar Ziabari, Singanallur V. Venkatakrishnan, Obaidullah Rahman, Gregery T. Buzzard, Charles A. Bouman

机构 * School of Electrical and Computer Engineering, Purdue University(普渡大学电气与计算机工程学院) ; Oak Ridge National Laboratory(橡树岭国家实验室)

AI总结 针对CT重建中稀疏视图选择的未解决问题,提出视图协方差损失(VCL)函数及对应的VCLS算法,实验表明该算法在固定视图数下的重建精度优于现有方法。

Comments Accepted version of the article published in IEEE Transactions on Pattern Analysis and Machine Intelligence. 11 pages main paper plus 3 pages supplementary material

Journal ref IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 48, no. 10, pp. 11335-11345, Oct. 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2611.07718 2026-11-06 cs.CV cs.LG 新提交

Multimodal Information Interactions Bottleneck for Multimodal Self-Supervised Learning

多模态自监督学习中的多模态信息交互瓶颈

Liangjian Wen, Jiangtao Zheng, Jiang Duan, Qun Dai, Yong Dai, Jianzhuang Liu, Jun Wang, Shu-Dong Huang, Xinwang Liu, Zenglin Xu, Zhao Kang

机构 * School of Computing and Artificial Intelligence, Southwestern University of Finance and Economics(西南财经大学计算机与人工智能学院) ; School of Management Science and Engineering, Southwestern University of Finance and Economics(西南财经大学管理科学与工程学院) ; Artificial Intelligence and Digital Finance Key Laboratory of Sichuan Province(四川省人工智能与数字金融重点实验室) ; Chengdu Everimaging Science and Technology Co., Ltd(成都影达科技有限公司) ; X-Humanoid Research Institute(类人X研究院) ; Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences(中国科学院深圳先进技术研究院) ; School of Artificial Intelligence, Sichuan University(四川大学人工智能学院) ; School of Computer, National University of Defense Technology(国防科技大学计算机学院) ; Artificial Intelligence Innovation and Incubation Institute, Fudan University(复旦大学人工智能创新与孵化研究院) ; Shanghai Academy of AI for Science(上海人工智能科学研究院) ; University of Electronic Science and Technology of China(电子科技大学)

AI总结 针对多模态自监督学习中对比方法忽略模态特异性与协同信息的问题,提出MIIB框架,通过双路径架构与熵目标优化,在多基准上实现性能提升。

Comments 15 pages, 3-page supplementary material provided as an ancillary file. Accepted for publication in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2611.07702 2026-11-06 cs.CV cs.AI cs.LG 新提交

InfMasking: A Unified Variational Framework for Synergistic Multimodal Learning

InfMasking:用于协同多模态学习的统一变分框架

Liangjian Wen, Qun Dai, Jiang Duan, Yong Dai, Jianzhuang Liu, Dongkai Wang, Jun Wang, Fuyuan Cao, Zenglin Xu, Zhao Kang

机构 * School of Computing and Artificial Intelligence, Southwestern University of Finance and Economics(西南财经大学计算机与人工智能学院) ; Chengdu Everimaging Science and Technology Co., Ltd.(成都亿美影科技有限公司) ; School of Management Science and Engineering, Southwestern University of Finance and Economics(西南财经大学管理科学与工程学院) ; Beijing Humanoid Robot Innovation Center (X-Humanoid)(北京人形机器人创新中心(X-Humanoid)) ; Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences(中国科学院深圳先进技术研究院) ; School of Computer and Information Technology, Shanxi University(山西大学计算机与信息技术学院) ; Artificial Intelligence Innovation and Incubation Institute, Fudan University(复旦大学人工智能创新与产业研究院) ; Shanghai Academy of AI for Science(上海人工智能科学研究院) ; University of Electronic Science and Technology of China(电子科技大学)

AI总结 InfMasking是基于无限掩码的多模态学习框架,通过掩码特征并约束融合表示一致性,抑制依赖主导模态,在基准上提升了预测性能与协同相关指标。

Comments Accepted for publication in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). 16 pages; supplementary material provided as an ancillary file. Code: https://github.com/brightest66/InfMasking

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2611.07124 2026-11-06 cs.LG 新提交

Machine Learning for Live-Birth Prediction After IVF/ICSI From Structured Clinical Data: A Systematic Review, Meta-Analysis, and Critical Appraisal

基于结构化临床数据的IVF/ICSI术后活产预测机器学习模型:系统评价、荟萃分析与批判性评估

Abdul Raheem Shahzad, Muhammad Zaheer Khan, Umar Ajaib, Asad Ur Rehman, Juan Deng, Ming Zhang

机构 * School of Computer Science, Wuhan University(武汉大学计算机学院) ; Reproductive Medical Center, Zhongnan Hospital of Wuhan University(武汉大学中南医院生殖医学中心) ; Clinical Medicine Research Center of Prenatal Diagnosis and Birth Health in Hubei Province(湖北省产前诊断与出生健康临床医学研究中心)

AI总结 本研究系统评价IVF/ICSI术后活产预测的机器学习模型,荟萃分析其性能,发现未明确决策点的模型AUC最高,模型提升源于分期预测与局部校准,而非深层架构,且多数研究偏倚风险高。

Comments 20 pages, 4 figures, 6 tables; supplementary material (26 pages) in ancillary files. Data and code: https://github.com/abdul-raheem-shahzad/ml-ivf-live-birth-review. Submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2611.06637 2026-11-05 cs.CV cs.AI 新提交

Human Motion Prediction via Continual Prior Compensation

基于持续先验补偿的人体运动预测

Jianwei Tang, Jian-Fang Hu, Tianming Liang, Xiaotong Lin, Jiangxin Sun, Wei-Shi Zheng, Jianhuang Lai

机构 * School of Computer Science, Sun Yat-sen University(中山大学计算机学院) ; Guangdong Province Key Laboratory of Information Security Technology(广东省信息安全技术重点实验室) ; Key Laboratory of Machine Intelligence and Advanced Computing (Sun Yat-sen University), Ministry of Education(教育部机器智能与先进计算重点实验室(中山大学))

AI总结 提出CPC与CPC++框架,将HMP任务划分为多子任务逐步训练,引入PCF/FGPCF缓解先验遗忘,在多数据集与骨干模型上验证了有效性与灵活性。

Journal ref IEEE Transactions on Pattern Analysis and Machine Intelligence (2026)

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2611.02769 2026-11-03 cs.CV 新提交

Unlocking the Potential of Vision Foundation Models for Multi-Task Dense Prediction via Mixture of Task Adapters

通过任务适配器混合模型释放视觉基础模型在多任务密集预测中的潜力

Yunzhi Zhuge, Xinzhuo Yu, Sitong Gong, Lu Zhang, Pingping Zhang, Huchuan Lu

机构 * School of Information and Communication Engineering, Dalian University of Technology(大连理工大学信息与通信工程学院) ; School of Computer Science and Technology, Dalian University of Technology(大连理工大学计算机科学与技术学院)

AI总结 该研究针对视觉基础模型多任务密集预测的冲突问题,提出D2MoE任务适配器混合模型,结合TG-LoRA与TA-MoE等模块,在少量可训练参数下实现了多任务性能提升,为相关感知任务提供了可扩展方案。

Comments Accepted to IEEE TPAMI

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2405.19005 2026-10-30 cs.CV 版本更新

AdalReID: Auto-selected Knowledge Adapters for Lifelong Person Re-identification

面向终身行人重识别的自动选择知识适配器

Xuelin Qian, Ruiqi Wu, Zhongling Huang, Dingwen Zhang, Junwei Han

机构 * Northwestern Polytechnical University(西北工业大学)

AI总结 针对终身行人重识别中的灾难性遗忘问题,提出采用知识适配器与无参数自动选择机制的AdalReID框架,通过逐步构建领域适配器并自适应选择与融合知识,在已见和未见领域上显著优于现有方法。

Comments Accepted by TPAMI

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.36487 2026-10-28 cs.CV 新提交

Spatial-Frequency Learning for Scanning Transmission Electron Microscopy Image Enhancement

用于扫描透射电子显微镜图像增强的空间-频率学习

Hesong Li, Ying Fu, Ziqi Wu, Ruiwen Shao, Tao Zhang

机构 * School of Medical Science and Engineering, Beijing Institute of Technology(北京理工大学医学科学与工程学院)

AI总结 针对严重退化下STEM图像增强性能有限的问题,提出空间-频率融合网络与STEM噪声校准方法,构建多样化数据集,实现了更优的图像增强效果与原子定位精度。

Comments Accepted by IEEE TPAMI

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.34461 2026-10-27 cs.CV 新提交

Task Relevance-Driven Prompt Expansion for Federated Continual Learning

面向联邦持续学习的任务相关性驱动提示扩展

Jiahua Dong, Wenqi Liang, Yang Cong, Yichen Li, Hongyuan Yu, Henghui Ding, Yulun Zhang, Salman Khan, Fahad Shahbaz Khan

机构 * Mohamed bin Zayed University of Artificial Intelligence(穆罕默德·本·扎耶德人工智能大学) ; University of Trento(特伦托大学) ; South China University of Technology(华南理工大学) ; Huazhong University of Science and Technology(华中科技大学) ; Xiaomi(小米) ; Institute of Big Data, Fudan University(复旦大学大数据研究院) ; Shanghai Jiao Tong University(上海交通大学)

AI总结 针对联邦持续学习依赖新任务先验知识、易受旧任务无关关系影响的问题,提出TRPE模型,通过分布外检测、自适应DBSCAN及全局任务相关性计算,缓解灾难性遗忘,实验验证其有效性。

Comments Accepted to Transactions on Pattern Analysis and Machine Intelligence (TPAMI)

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2608.09438 2026-10-27 cs.CV 版本更新

Unveiling the Secret of AdaLN-Zero in Diffusion Transformer

揭示扩散Transformer中AdaLN-Zero的秘密

Jie Zhu, Mingyu Ding, Boqiang Duan, Leye Wang, Jingdong Wang

机构 * Peking University(北京大学) ; UC Berkeley(加州大学伯克利分校) ; Baidu(百度)

AI总结 本研究探究扩散Transformer(DiT)中AdaLN-Zero性能优于AdaLN的原因,发现零初始化是关键要素,提出AdaLN-Gaussian初始化策略与SE-adaLN-Zero机制,经多数据集实验验证其有效性与泛化性。

Comments Accepted by IEEE TPAMI 2026, camera-ready version

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.21539 2026-10-26 cs.AI 版本更新

Reinforcement Learning-driven Diffusion Models: Language, Vision, and Beyond

强化学习驱动的扩散模型:语言、视觉及其他领域

Zhiyuan Ma, Jiabao Wei, Tianyu Wang, Hanqing Wang, Jiadong Pan, Haowen Xu, Liangliang Zhao, Ruixun Liu, Yuhan Wang, Yuwei Zhang, Liang Feng, Junhao Xiao, Kaiyan Zhang, Xuekai Zhu, Daizong Liu, Biqing Qi, Yihao Liu, Ganqu Cui, Ning Ding, Linfeng Zhang, Jianjun Li, Kun He, Xiang Bai, Bowen Zhou

机构 * Huazhong University of Science and Technology(华中科技大学) ; Beijing Institute of Technology(北京理工大学) ; Tsinghua University(清华大学) ; Hong Kong University of Science and Technology(香港科技大学) ; University of Chinese Academy of Sciences(中国科学院大学) ; Fudan University(复旦大学) ; Shanghai AI Laboratory(上海人工智能实验室) ; Peking University(北京大学) ; Tongji University(同济大学) ; Wuhan University(武汉大学)

AI总结 本文综述RL驱动的扩散模型,分析其在语言、视觉等领域的作用,评估相关方法优势,剖析部署瓶颈并给出路线图与未来方向。

Comments 26 pages,12 figures,8 table,accepted by TPAMI

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.10959 2026-10-21 cs.CV 版本更新

GPU-Accelerated Computation of Persistent Homology for Topological Analysis of Image Data

用于图像数据拓扑分析的持续同调的GPU加速计算

Fan Wang, Hubert Wagner, Rezaul Chowdhury, Chao Chen

机构 * Stony Brook University(石溪大学) ; University of Florida(佛罗里达大学)

AI总结 本文提出GPU加速的TopoGPU流管道,结合分层感知离散Morse匹配等技术,实现持续同调计算的大幅提速,优于Cubical Ripser,集成入深度网络可降低训练成本,为开源项目

Comments Accepted for publication in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). Includes supplementary material (8 pages) after the main paper. Code: https://github.com/seravee08/GPU-Computation-of-Persistent-Homology-for-Image-Data

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2604.28196 2026-10-21 cs.CV 版本更新

HERMES++: Toward a Unified Driving World Model for 3D Scene Understanding and Generation

HERMES++:迈向统一的驾驶世界模型用于3D场景理解和生成

Xin Zhou, Dingkang Liang, Xiwu Chen, Feiyang Tan, Dingyuan Zhang, Hengshuang Zhao, Xiang Bai

机构 * Huazhong University of Science and Technology(华中科技大学) ; Mach Drive ; University of Hong Kong(香港大学)

AI总结 本文提出HERMES++,一种统一的驾驶世界模型,整合3D场景理解和未来几何预测。通过BEV表示、LLM增强世界查询和当前到未来链接等设计,提升驾驶场景的生成与理解能力。

Comments Accepted by IEEE TPAMI. Code: https://github.com/H-EmbodVis/HERMESV2, Project page: https://h-embodvis.github.io/HERMESV2/

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2511.15308 2026-10-20 cs.CV 版本更新

Text2Loc++: Generalizing 3D Point Cloud Localization from Natural Language

Text2Loc++:基于自然语言的3D点云定位泛化方法

Yan Xia, Letian Shi, Yilin Di, Joao F. Henriques, Daniel Cremers

机构 * School of Artificial Intelligence and Data Science, University of Science and Technology of China(人工智能与数据科学学院,中国科学技术大学) ; Technical University of Munich(慕尼黑技术大学) ; Visual Geometry Group, University of Oxford(牛津大学视觉几何组)

AI总结 针对自然语言描述下的3D点云子地图定位问题,提出由粗到精的跨模态对齐网络Text2Loc++,构建新城市级数据集,在KITTI360Pose上性能较现有方法最高提升15%且泛化性强。

Comments This paper has been published in IEEE-TPAMI 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2410.08524 2026-10-20 cs.LG

IGNN-Solver: A Graph Neural Solver for Implicit Graph Neural Networks

IGNN-Solver:一种用于隐式图神经网络的图神经求解器

Junchao Lin, Zenan Ling, Zhanbo Feng, Jingwen Xu, Minxuan Liao, Feng Zhou, Tianqi Hou, Zhenyu Liao, Robert C. Qiu

机构 * School of EIC, Huazhong University of Science and Technology(华中科技大学电子信息与通信学院) ; Shanghai Jiao Tong University(上海交通大学) ; Renmin University of China(中国人民大学) ; Huawei(华为) ; School of Science, Wuhan University of Technology(武汉理工大学理学院)

AI总结 针对隐式图神经网络定点迭代成本高、难扩展的问题,提出由微型GNN参数化的IGNN-Solver,并结合稀疏化与存储压缩,在保持精度下实现1.5–8倍推理加速。

Journal ref IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 1-18, 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2504.00470 2026-10-19 cs.LG cs.CV 版本更新

Less is More: Efficient Black-box Attribution via Minimal Interpretable Subset Selection

少即是多:通过最小可解释子集选择实现高效的黑盒属性分析

Ruoyu Chen, Siyuan Liang, Jingzhi Li, Shiming Liu, Li Liu, Hua Zhang, Xiaochun Cao

机构 * Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所) ; University of Chinese Academy of Sciences(中国科学院大学) ; College of Computing and Data Science, Nanyang Technological University(南洋理工大学计算机与数据科学学院) ; School of Artificial Intelligence, University of Science and Technology Beijing(北京科技大学人工智能学院) ; Department of Mechanical Engineering, Imperial College London(伦敦帝国理工学院机械工程系) ; Center for Machine Vision and Signal Analysis (CMVS), University of Oulu(奥卢大学机器视觉与信号分析中心) ; School of Cyber Science and Technology, Shenzhen Campus of Sun Yat-sen University(中山大学深圳校区计算机科学与技术学院)

AI总结 本文提出了一种高效的黑盒属性分析方法LiMA,通过将重要区域的属性分析转化为子模函数子集选择的优化问题,以更少的区域提供更准确的解释,并在多个基准模型上展示了显著的改进。

Comments Accepted by T-PAMI (IEEE Transactions on Pattern Analysis and Machine Intelligence)

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.21804 2026-10-16 cs.AI 新提交

Co2In: a Bi-level Memory Incremental Learning Framework with Knowledge Encoding, Consolidation, and Integration

Co2In:一种具有知识编码、巩固与整合功能的双层记忆增量学习框架

Wenju Sun, Qingyong Li, Boyang Li, Wen Wang, Siyu Zhang, Yi Wei, Yangliao Geng

机构 * Beijing Jiaotong University(北京交通大学) ; Nanyang Technological University(南洋理工大学) ; Wuyi University(武夷学院)

AI总结 受人类记忆机制启发,提出双层记忆架构的Co2In框架,通过编码、巩固、整合实现增量学习的可塑性与稳定性平衡,在三类数据集的两种设置下取得高效内存的最优性能。

Comments Accepted by IEEE Transactions On Pattern Analysis and Machine Intelligence

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2412.12877 2026-10-15 cs.CV 版本更新

PRIMEdit: Probability Redistribution for Instance-Aware Multi-Object Video Editing With a New Benchmark Dataset

PRIMEdit:面向实例感知多对象视频编辑的概率重分配与基准数据集

Samuel Teodoro, Agus Gunawan, Soo Ye Kim, Jihyong Oh, Munchurl Kim

机构 * KAIST(韩国科学技术院) ; Adobe Research(奥多比研究院) ; Chung-Ang University(中央大学)

AI总结 提出零样本框架PRIMEdit,通过实例中心概率重分配和分离多实例采样,解决多对象视频编辑中的不忠实与泄漏问题,并引入MIVE数据集和CIA指标,显著优于现有方法。

Comments Accepted manuscript published in IEEE TPAMI, vol. 48, no. 11, pp. 15571-15588, Nov. 2026. The first two authors contributed equally to this work. The last two authors are co-corresponding authors. Please visit our project page at https://kaist-viclab.github.io/primedit-site/

Journal ref IEEE Trans. Pattern Anal. Mach. Intell., vol. 48, no. 11, pp. 15571-15588, Nov. 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2601.02759 2026-10-14 cs.CV cs.RO 版本更新

Toward Zero-Shot Point Cloud Registration Across Diverse Scales, Scenes, and Sensor Setups

跨多样尺度、场景和传感器设置的零样本点云配准

Hyungtae Lim, Minkyun Seo, Luca Carlone, Jaesik Park

机构 * Laboratory for Information and Decision Systems (LIDS), Massachusetts Institute of Technology(信息与决策系统实验室,麻省理工学院) ; Department of Computer Science and Engineering, Seoul National University(计算机科学与工程系,首尔国立大学)

AI总结 提出BUFFER-X框架,通过几何自举、分布感知采样和坐标归一化实现零样本点云配准,同时引入BUFFER-X-Lite提升效率,适用于多样场景和传感器设置。

Comments 23 pages, 21 figures, TPAMI 2026. Extended version of our ICCV 2025 highlight paper [arXiv:2503.07940]. arXiv admin note: substantial text overlap with arXiv:2503.07940

Journal ref IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 48, no. 11, pp. 15234-15251, Nov. 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.12095 2026-10-09 cs.CV cs.LG 新提交

DVLA-RL++: Dual-Level Vision-Language Alignment with Reinforcement Learning Gating for Few-Shot Learning

DVLA-RL++:用于小样本学习的带强化学习门控的双层级视觉-语言对齐方法

Wenhao Li, Xianjing Meng, Qiangchang Wang, Zhongyi Han, Yilong Yin, Liqiang Nie

机构 * School of Software, Shandong University(山东大学软件学院) ; Shenzhen Loop Area Institute(深圳河套学院) ; School of Computer and Artificial Intelligence, Shandong University of Finance and Economics(山东财经大学计算机与人工智能学院) ; School of Computer Science and Technology, Harbin Institute of Technology (Shenzhen)(哈尔滨工业大学(深圳)计算机科学与技术学院)

AI总结 本文针对小样本学习中支持原型易受上下文污染的问题,提出带互补语义纯化与反事实强化学习门控的DVLA-RL++,在多类基准上较DVLA-RL平均提升1.4%准确率,达到当前最优性能。

Comments This work has been submitted to the IEEE TPAMI for possible publication

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.11920 2026-10-09 cs.CL 新提交

Event-Centric Memory with Query-Aware Graph Augmentation for Long-Term Conversational Agents

面向长期对话智能体的、带有查询感知图增强的以事件为中心的记忆

Yichen Liu, Chunfeng Yuan, Haowei Liu, Wenjuan Li, Zefeng Lin, Bing Li, Xu Chen, Weiming Hu

机构 * Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所) ; University of Chinese Academy of Sciences(中国科学院大学) ; Alibaba Group(阿里巴巴集团) ; PeopleAI Inc.(北京人人智能科技有限公司) ; Renmin University of China(中国人民大学) ; ShanghaiTech University(上海科技大学)

AI总结 针对现有对话智能体记忆系统的局限,提出 QGMem 框架,通过以事件为中心的记忆组织与查询感知图增强,在六个基准任务中实现了检索、多跳推理等任务的性能提升。

Comments 17 pages, 7 figures. Submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.11313 2026-10-09 cs.LG 新提交

NP-Hardness of Minimizing Neurons in Two-Hidden-Layer ReLU Neural Networks

两层隐藏层ReLU神经网络中最小化神经元数量的NP-难解性

Sangrock Lee

机构 * FAMU–FSU College of Engineering(佛罗里达农工大学-佛罗里达州立大学工程学院) ; Florida State University(佛罗里达州立大学)

AI总结 本文证明两层隐藏层ReLU网络中最小化隐藏神经元数量的问题是NP-难解的,为ReLU网络设计采用启发式逼近方法提供了理论依据。

Journal ref Lee, S., NP-Hardness of Minimizing Neurons in Two-Hidden-Layer ReLU Neural Networks, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 48, no. 11, 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2504.12392 2026-10-09 stat.ME cs.LG stat.ML 版本更新

A Survey on Archetypal Analysis

关于原型分析的综述

Aleix Alcacer, Irene Epifanio, Sebastian Mair, Morten Mørup

机构 * Jaume I University(杰拉尔·伊大学) ; ValgrAI(瓦尔格AI) ; Linköping University(林雪平大学) ; Technical University of Denmark(丹麦技术大学)

AI总结 本文综述了原型分析的方法、应用及未来研究方向,强调其在数据建模中的优势与挑战。

Comments 28 pages, 14 figures, accepted at TPAMI

Journal ref IEEE Transactions on Pattern Analysis and Machine Intelligence 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2610.09579 2026-10-08 cs.CV 新提交

Gradient-Based Trajectory Optimisation over Continuous Poses for Sparse-View Cone-Beam CT

基于连续位姿梯度的稀疏视角锥束CT轨迹优化

Linda-Sophie Schneider, Simon Wittl, Gabriel Herl, Andreas Maier

机构 * Friedrich-Alexander-Universität Erlangen-Nürnberg(埃尔朗根-纽伦堡弗里德里希-亚历山大大学) ; Deggendorf Institute of Technology(德根多夫技术学院)

AI总结 提出将稀疏视角锥束CT的每个源位姿作为连续变量,通过梯度上升联合优化轨迹,集成软Tuy覆盖、视角协方差损失与衰减惩罚,在多种扫描几何上提升缺陷可见性并支持快速预扫描规划。

Comments Submitted to TPAMI

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2603.19899 2026-10-07 stat.ML cs.LG stat.AP 版本更新

Deep Time-Series Forecasting in 10 Years: A Survey

深度自相关建模用于时间序列预测:进展与展望

Hao Wang, Licheng Pan, Qingsong Wen, Jialin Yu, Zhichao Chen, Chunyuan Zheng, Xiaoxi Li, Zhixuan Chu, Chao Xu, Mingming Gong, Haoxuan Li, Yuan Lu, Zhouchen Lin, Philip Torr, Yan Liu

AI总结 本文从自相关建模角度综述了深度时间序列预测的进展,提出新的分类体系并分析了模型架构与学习目标的演进,填补了现有研究的空白。

Comments This survey is accepted by IEEE TPAMI

Journal ref IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2502.19741 2026-10-07 cs.LG 版本更新

Causal Effect Estimation under Networked Interference without Networked Unconfoundedness Assumption

在不假设网络无偏置的情况下估计网络干扰的因果效应

Weilin Chen, Ruichu Cai, Jie Qiao, Yuguang Yan, José Miguel Hernández-Lobato

机构 * Guangdong University of Technology(广东工业大学) ; Peng Cheng Laboratory(鹏城实验室) ; University of Cambridge(剑桥大学)

AI总结 本文提出了一种无需网络无偏置假设的网络效应估计方法,通过恢复潜在混杂因素来提高因果效应估计的准确性。

Comments accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence, in press

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
2510.06842 2026-10-06 cs.CV 版本更新

Continual Action Quality Assessment via Adaptive Manifold-Aligned Graph Regularization

持续动作质量评估:基于自适应流形对齐图正则化

Kanglei Zhou, Qingyi Pan, Xingxing Zhang, Hubert P. H. Shum, Frederick W. B. Li, Xiaohui Liang, Liyuan Wang

机构 * Department of Psychological and Cognitive Sciences, Tsinghua University(心理与认知科学系,清华大学) ; Institute for AI, BNRist Center, Tsinghua-Bosch Joint ML Center, THBI Lab, Tsinghua University(人工智能研究院、北京人工智能研究所、清华大学-博世联合机器学习中心、THBI实验室,清华大学) ; Department of Computer Science, Durham University(计算机科学系,杜尔姆大学) ; State Key Laboratory of Virtual Reality Technology and Systems, Beihang University(虚拟现实技术与系统国家重点实验室,北京航空航天大学) ; Zhongguancun Laboratory, Beijing(中关村实验室,北京)

AI总结 针对动作质量评估中分布非平稳问题,提出持续学习框架CAQA及自适应流形对齐图正则化MAGR++,通过全参数微调与特征校正,在四个基准上平均相关性提升3.6%离线、12.2%在线。

Comments Accepted to IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)

Journal ref IEEE Transactions on Pattern Analysis and Machine Intelligence, 2026

展开摘要

展开后加载摘要…

URL PDF HTML 收藏
↑