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)

2025-12-23 至 2025-12-23 共收录 13
2412.11154 2025-12-23 cs.CV

From Easy to Hard: Progressive Active Learning Framework for Infrared Small Target Detection with Single Point Supervision

从易到难:基于单点监督的红外小目标检测渐进主动学习框架

Chuang Yu, Jinmiao Zhao, Yunpeng Liu, Sicheng Zhao, Yimian Dai, Xiangyu Yue

机构 * Key Laboratory of Opto-Electronic Information Processing, Chinese Academy of Sciences(光电信息处理重点实验室,中国科学院) Shenyang Institute of Automation, Chinese Academy of Sciences(沈阳自动化研究所,中国科学院) University of Chinese Academy of Sciences(中国科学院大学) Tsinghua University(清华大学) Nankai University(南开大学) MMLab, The Chinese University of Hong Kong(香港中文大学MMLab) CPII under InnoHK(创新香港下的CPII)

AI总结 本文提出渐进主动学习框架,通过模型预启动和双更新策略提升单点监督下红外小目标检测的性能和稳定性。

Comments Accepted by ICCV 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.19271 2025-12-23 cs.CV

3SGen: Unified Subject, Style, and Structure-Driven Image Generation with Adaptive Task-specific Memory

3SGen: 一种统一的主体、风格和结构驱动的图像生成方法,具有自适应任务特定记忆

Xinyang Song, Libin Wang, Weining Wang, Zhiwei Li, Jianxin Sun, Dandan Zheng, Jingdong Chen, Qi Li, Zhenan Sun

机构 * School of Artificial Intelligence, UCAS(人工智能学院,UCAS) CASIA AntGroup(蚂蚁集团)

AI总结 3SGen通过统一的框架实现主体、风格和结构驱动的图像生成,采用自适应任务特定记忆模块提升生成质量和跨任务迁移性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.19081 2025-12-23 cs.AI

Population-Evolve: a Parallel Sampling and Evolutionary Method for LLM Math Reasoning

Population-Evolve: 一种用于LLM数学推理的并行采样和进化方法

Yanzhi Zhang, Yitong Duan, Zhaoxi Zhang, Jiyan He, Shuxin Zheng

机构 * Academy of Mathematics and Systems Science, Chinese Academy of Sciences(中国科学院数学与系统科学研究院) University of Chinese Academy of Sciences(中国科学院大学) Zhongguancun Academy(中关村学院) Zhongguancun Institute of Artificial Intelligence(中关村人工智能研究院)

AI总结 Population-Evolve通过并行推理和进化提示提升LLM数学推理能力,实现高准确性和低计算成本。

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.18651 2025-12-23 cs.CV

Adversarial Robustness in Zero-Shot Learning:An Empirical Study on Class and Concept-Level Vulnerabilities

零样本学习中的对抗鲁棒性:对类别和概念层面脆弱性的实证研究

Zhiyuan Peng, Zihan Ye, Shreyank N Gowda, Yuping Yan, Haotian Xu, Ling Shao

机构 * iFLYTEK Co., Ltd.(iFLYTEK公司) UCAS-Terminus AI Lab, University of Chinese Academy of Sciences(中国科学院大学Terminus AI实验室) School of Computer Science, the University of Nottingham(诺丁汉大学计算机学院) TGAI lab, the Westlake University(西湖大学TGAI实验室) RippleInfo Co., Ltd(RippleInfo公司)

AI总结 本研究通过实证分析揭示了零样本学习模型在类别和概念层面的对抗脆弱性,并提出了改进对抗鲁棒性的方法。

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.17370 2025-12-23 cs.RO cs.AI

TakeAD: Preference-based Post-optimization for End-to-end Autonomous Driving with Expert Takeover Data

TakeAD: 基于偏好的端到端自动驾驶后优化方法与专家接管数据

Deqing Liu, Yinfeng Gao, Deheng Qian, Qichao Zhang, Xiaoqing Ye, Junyu Han, Yupeng Zheng, Xueyi Liu, Zhongpu Xia, Dawei Ding, Yifeng Pan, Dongbin Zhao

机构 * The 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,中国科学院自动化研究所) 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 Automation and Electrical Engineering, University of Science and Technology Beijing(北京科技大学自动化与电气工程学院) Chongqing Chang’an Technology Co., Ltd.(重庆长安科技有限公司)

AI总结 TakeAD通过基于偏好的后优化框架利用专家接管数据,提升端到端自动驾驶闭环性能。

Comments This work has been accepted by IEEE RA-L. Manuscript submitted: July, 8, 2025; Accepted: November, 24, 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2504.12826 2025-12-23 cs.RO cs.CV

UncAD: Towards Safe End-to-end Autonomous Driving via Online Map Uncertainty

UncAD: 向通过在线地图不确定性实现安全端到端自动驾驶迈进

Pengxuan Yang, Yupeng Zheng, Qichao Zhang, Kefei Zhu, Zebin Xing, Qiao Lin, Yun-Fu Liu, Zhiguo Su, Dongbin Zhao

机构 * Key Laboratory of Safety Intelligent Mining in Non-coal Open-pit Mines, National Mine safety Administration, Guangdong Guangzhou, 510000, China(安全智能采矿非煤矿山重点实验室,国家矿山安全监察局,广东广州,510000,中国) The 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, Beijing, China(人工智能学院,中国科学院大学,北京,中国) EACON, Fujian, China(福建中国EACON)

AI总结 UncAD通过引入在线地图不确定性,提升自动驾驶安全性,减少碰撞和冲突率。

Journal ref 2025 IEEE International Conference on Robotics and Automation (ICRA)

详情

展开后加载摘要…

URL PDF HTML 收藏
2503.11185 2025-12-23 cs.CR cs.AI

Bleeding Pathways: Vanishing Discriminability in LLM Hidden States Fuels Jailbreak Attacks

出血路径:LLM隐藏状态中的判别能力消失加剧了 jailbreak 攻击

Yingjie Zhang, Tong Liu, Zhe Zhao, Guozhu Meng, Kai Chen

机构 * Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所) School of Cyber Security, University of Chinese Academy of Sciences(中国科学院大学网络安全学院) Ant Group(蚂蚁集团)

AI总结 本研究提出DEEPALIGN框架,通过增强LLM隐藏状态的分离度,有效缓解jailbreak攻击,提升安全性和实用性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2501.17642 2025-12-23 cs.CV

Efficient Redundancy Reduction for Open-Vocabulary Semantic Segmentation

高效开放词汇语义分割中的冗余减少

Lin Chen, Qi Yang, Kun Ding, Zhihao Li, Gang Shen, Fei Li, Qiyuan Cao, Shiming Xiang

机构 * State Key Laboratory of Multimodal Artificial Intelligence Systems (MAIS), Institute of Automation, Chinese Academy of Sciences(多模态人工智能系统国家重点实验室(MAIS)、自动化研究所、中国科学院) 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 Software, Shandong University(山东大学软件学院) China Tower Corporation Limited(中国铁塔股份有限公司)

AI总结 本文提出ERR-Seg,通过减少冗余信息和优化序列建模,提升开放词汇语义分割的效率与性能。

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.18571 2025-12-23 cs.AI cs.CV

ESearch-R1: Learning Cost-Aware MLLM Agents for Interactive Embodied Search via Reinforcement Learning

ESearch-R1: 通过强化学习学习成本感知的多模态大语言模型代理以进行交互式具身搜索

Weijie Zhou, Xuangtang Xiong, Ye Tian, Lijun Yue, Xinyu Wu, Wei Li, Chaoyang Zhao, Honghui Dong, Ming Tang, Jinqiao Wang, Zhengyou Zhang

机构 * School of Traffic and Transportation, Beijing Jiaotong University(交通与运输学院,北京交通大学) Tencent Robotics X & Futian Laboratory(腾讯机器人X与福田实验室) Foundation Model Research Center, Institute of Automation, Chinese Academy of Sciences(基础模型研究中心,中国科学院自动化研究所) University of Chinese Academy of Sciences(中国科学院大学)

AI总结 ESearch-R1通过强化学习方法,结合交互对话、记忆检索和导航,实现成本感知的多模态大语言模型代理,有效降低任务执行成本并提高成功率。

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.18411 2025-12-23 cs.CV cs.AI

AmPLe: Supporting Vision-Language Models via Adaptive-Debiased Ensemble Multi-Prompt Learning

AmPLe: 通过自适应去偏集成多提示学习支持视觉-语言模型

Fei Song, Yi Li, Jiangmeng Li, Rui Wang, Changwen Zheng, Fanjiang Xu, Hui Xiong

机构 * National Key Laboratory of Space Integrated Information System, Institute of Software, Chinese Academy of Sciences(中国科学院空间信息集成系统国家重点实验室,软件研究所) University of Chinese Academy of Sciences(中国科学院大学) The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州)) The Hong Kong University of Science and Technology(香港科技大学)

AI总结 AmPLe通过自适应去偏集成多提示学习方法,解决模型-提示匹配偏差和样本-提示匹配偏差,提升视觉-语言模型在下游任务中的性能。

Comments Accepted by IJCV2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.18251 2025-12-23 cond-mat.mtrl-sci cs.LG physics.comp-ph

CrystalFormer-CSP: Thinking Fast and Slow for Crystal Structure Prediction

CrystalFormer-CSP: 快速与缓慢思考的晶体结构预测

Zhendong Cao, Shigang Ou, Lei Wang

机构 * Institute of Physics, Chinese Academy of Sciences, Beijing, China(中国科学院物理研究所) School of Physics, University of Chinese Academy of Sciences, Beijing, China(中国科学院大学物理学院)

AI总结 CrystalFormer-CSP结合数据驱动和物理驱动方法,通过预训练生成模型和力场优化,高效预测稳定晶体结构,并通过强化微调提升准确性。

Comments 11 pages, 4 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2506.07576 2025-12-23 cs.CV

Super Encoding Network: Recursive Association of Multi-Modal Encoders for Video Understanding

超级编码网络:多模态编码器的递归关联用于视频理解

Boyu Chen, Siran Chen, Kunchang Li, Qinglin Xu, Yu Qiao, Yali Wang

机构 * Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences(中国科学院深圳先进技术研究院) the School of Artificial Intelligence, University of Chinese Academy of Sciences(中国科学院大学人工智能学院) Shanghai AI Laboratory(上海人工智能实验室)

AI总结 本文提出超级编码网络,通过递归关联多模态编码器提升视频理解性能,显著提升跟踪、识别、聊天和编辑等任务效果。

详情

展开后加载摘要…

URL PDF HTML 收藏
2307.14596 2025-12-23 cs.LG

HUTFormer: Hierarchical U-Net Transformer for Long-Term Traffic Forecasting

HUTFormer:分层U-Net变换器用于长期交通预测

Zezhi Shao, Fei Wang, Tao Sun, Chengqing Yu, Yuchen Fang, Guangyin Jin, Zhulin An, Yang Liu, Xiaobo Qu, Yongjun Xu

机构 * Institute of Computing Technology, Chinese Academy of Sciences(中国科学院计算技术研究所) School of Computer Science and Technology, University of Chinese Academy of Sciences(中国科学院大学计算机科学与技术学院) School of Computer Science and Engineering, University of Electronic Science and Technology of China(电子科技大学计算机科学与工程学院) Department of Planning, Design, and Technology of Architecture, Sapienza University of Rome(罗马大学建筑规划、设计与技术系) School of Vehicle and Mobility, Tsinghua University(清华大学车辆与移动性学院)

AI总结 HUTFormer通过分层U-Net变换器解决长期交通预测问题,采用多尺度表示和高效嵌入策略提升预测性能。

Comments Accepted for publication in the Communications in Transportation Research on December 2025. 38 paqes with 9 fiqures and 6 tables

Journal ref Communications in Transportation Research. 5, 100218 (2025)

详情

展开后加载摘要…

URL PDF HTML 收藏