arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

Zhejiang University(浙江大学)

2026-06-02 至 2026-06-02 共收录 45
2603.00171 2026-06-02 cs.CV cs.AI

LookWise: Knowing When and Where to Look for Fine-Grained Visual Reasoning in Multimodal Large Language Models

LookWise: 知道何时何地关注多模态大语言模型中的细粒度视觉推理

Yuxiang Shen, Hailong Huang, Zhenkun Gao, Xueheng Li, Man Zhou, Chengjun Xie, Haoxuan Che, Xuanhua He, Jie Zhang

机构 * Institute of Intelligent Machines, Hefei Institutes of Physical Science, Chinese Academy of Sciences(智能机器研究所,合肥物理科学研究院,中国科学院) University of Science and Technology of China(中国科学技术大学) Zhejiang University(浙江大学) East China Normal University(华东师范大学) The Hong Kong University of Science and Technology(香港科技大学)

AI总结 提出LookWise框架,通过置信度模块和语义引导定位模块实现自适应视觉推理,无需额外训练即可提升细粒度推理精度并加速推理。

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.03202 2026-06-02 cs.CL

Code2Math: Can Your Code Agent Effectively Evolve Math Problems Through Exploration?

Code2Math:你的代码智能体能否通过探索有效演化数学问题?

Dadi Guo, Yuejin Xie, Qingyu Liu, Weixian Huang, Jiayu Liu, Zhiyuan Fan, Qihan Ren, Shuai Shao, Tianyi Zhou, Jianjie Feng, Wenze Su, Yujiu Yang, Dongrui Liu, Yi R. Fung

机构 * Hong Kong University of Science and Technology(香港科技大学) Tsinghua University(清华大学) Zhejiang University(浙江大学) Nanjing Tech University(南京工业大学) Shanghai Jiao Tong University(上海交通大学) University of Michigan(密歇根大学) Independent Researcher(独立研究者)

AI总结 本文提出一个多智能体框架,利用代码智能体通过探索将现有数学问题自主演化为更复杂、更困难的变体,并验证其可解性和难度提升。

Comments 38 pages

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.17706 2026-06-02 cs.LG

Parallel Complex Diffusion for Scalable Time Series Generation

并行复数扩散用于可扩展时间序列生成

Rongyao Cai, Yuxi Wan, Kexin Zhang, Ming Jin, Zhiqiang Ge, Qingsong Wen, Yong Liu

机构 * Institute of Cyber-Systems and Control(网络系统与控制研究所) Zhejiang University(浙江大学) Griffith University(格里菲斯大学) School of Mathematics(数学学院) Southeast University(东南大学) Squirrel Ai Learning

AI总结 提出PaCoDi(并行复数扩散)框架,通过离散傅里叶变换将时间序列分解到谱域,利用并行实值估计器替代复数估计器,解决时间序列生成中的纠缠问题,理论证明谱高斯噪声的正交性,并引入平均场理论近似处理边缘耦合,在无条件和条件生成任务中优于5个基线。

Comments Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD '26). Extended Version with Full Proofs

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.10623 2026-06-02 cs.LG cs.AI

Mitigating Reward Hacking in RLHF via Bayesian Non-negative Reward Modeling

通过贝叶斯非负奖励建模缓解RLHF中的奖励黑客

Zhibin Duan, Guowei Rong, Zhuo Li, Bo Chen, Mingyuan Zhou, Dandan Guo

机构 * Zhejiang University(浙江大学)

AI总结 提出贝叶斯非负奖励模型(BNRM),通过非负因子分析和变分推断,在Bradley-Terry偏好模型中实现解耦与去偏,有效缓解奖励过度优化,提升鲁棒性和可解释性。

Comments Accepted as an Oral presentation at ICML 2026. The code is available at https://github.com/GuoweiRong/Bayesian-Non-negative-Reward-Model

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.00753 2026-06-02 math.OC cs.LG

Mirror Descent Under Generalized Smoothness

广义光滑性下的镜像下降

Dingzhi Yu, Wei Jiang, Hongyi Tao, Yuanyu Wan, Lijun Zhang

机构 * State Key Laboratory of Novel Software Technology, Nanjing University(南京大学新型软件技术国家重点实验室) School of Artificial Intelligence, Nanjing University(南京大学人工智能学院) School of Computer Science and Engineering, Nanjing University of Science and Technology(南京理工大学计算机科学与工程学院) School of Software Technology, Zhejiang University(浙江大学软件技术学院) Hangzhou High-Tech Zone (Binjiang) Institute of Blockchain and Data Security(杭州高新技术区(滨江)区块链与数据安全研究院)

AI总结 本文提出一种新的 $\ell_*$-光滑性概念,将经典光滑性推广到一般范数空间,并证明镜像下降类算法在此条件下收敛率与经典光滑性一致。

Comments ICML 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.08948 2026-06-02 cs.IR cs.AI

SHERLOCK: Towards Dynamic Knowledge Adaptation in LLM-enhanced E-commerce Risk Management

SHERLOCK:面向LLM增强电商风险管理的动态知识适应

Nan Lu, Yurong Hu, Jiaquan Fang, Yan Liu, Rui Dong, Yiming Wang, Rui Lin, Shaoyi Xu

机构 * Beijing Jiaotong University(北京交通大学) JD.com(京东公司) Southeast University(东南大学) Zhejiang University(浙江大学)

AI总结 提出Sherlock框架,通过构建领域知识库、两阶段检索增强生成和自演化数据飞轮,将结构化知识与LLM推理结合,提升电商风险案例调查的效率和准确性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.06448 2026-06-02 cs.LG cs.AI

Principle-Evolvable Scientific Discovery via Uncertainty Minimization

通过不确定性最小化实现原理可演化的科学发现

Yingming Pu, Tao Lin, Hongyu Chen

机构 * Westlake University(西lake大学) Zhejiang University(浙江大学)

AI总结 提出PiEvo框架,将科学发现视为原理空间上的贝叶斯优化,通过信息导向假设选择与异常驱动增强机制,使智能体自主演化理论世界观,在四个基准上平均解质量达90.81%~93.15%,收敛速度提升83.3%。

Comments Proceedings of the 43rd International Conference on Machine Learning, Seoul, South Korea. PMLR 306, 2026. Copyright 2026 by the author(s)

Journal ref Proc. 43rd Intl. Conf. on Machine Learning (ICML 2026), PMLR 306

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.03203 2026-06-02 cs.CL cs.LG

ForesightKV: Optimizing KV Cache Eviction for Reasoning Models by Learning Long-Term Contribution

ForesightKV: 通过学习长期贡献优化推理模型的KV缓存驱逐

Zican Dong, Peiyu Liu, Junyi Li, Zhipeng Chen, Han Peng, Shuo Wang, Wayne Xin Zhao

机构 * Zhejiang University(浙江大学)

AI总结 提出ForesightKV框架,通过监督学习和强化学习预测KV对在长文本生成中的重要性,在仅用一半缓存预算下优于现有方法。

Comments ICML 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.22965 2026-06-02 cs.RO

Self-Imitated Diffusion Policy for Efficient and Robust Visual Navigation

自模仿扩散策略用于高效鲁棒的视觉导航

Runhua Zhang, Junyi Hou, Changxu Cheng, Qiyi Chen, Tao Wang, Wuyue Zhao

机构 * Uni-Ubi Technology Co., Ltd.(Uni-Ubi技术有限公司) College of Control Science and Engineering, Zhejiang University(浙江大学控制科学与工程学院)

AI总结 提出自模仿扩散策略(SIDP),通过奖励引导的自模仿机制和课程学习范式,减少对大量采样和后过滤的依赖,实现高效鲁棒的视觉导航。

Comments Preprint

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.17502 2026-06-02 cs.RO

RynnVLA-002: A Unified Vision-Language-Action and World Model

RynnVLA-002: 统一的视觉-语言-动作与世界模型

Jun Cen, Siteng Huang, Yuqian Yuan, Kehan Li, Hangjie Yuan, Chaohui Yu, Bohan Hou, Yuming Jiang, Jiayan Guo, Xin Li, Hao Luo, Fan Wang, Deli Zhao, Hao Chen

机构 * DAMO Academy, Alibaba Group(达摩院,阿里巴巴集团) Hupan Lab(虎扑实验室) Zhejiang University(浙江大学)

AI总结 提出统一视觉-语言-动作(VLA)与世界模型的框架RynnVLA-002,通过联合学习环境动态和动作规划,在仿真和真实机器人任务中显著提升成功率。

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.07910 2026-06-02 cs.CL

Last Layer Logits to Logic: Empowering LLMs with Logic-Consistent Structured Knowledge Reasoning

最后一层逻辑到逻辑:赋予LLM逻辑一致的结构化知识推理能力

Songze Li, Zhiqiang Liu, Zhaoyan Gong, Xiaoke Guo, Zhongpu Bo, Zhengke Gui, Lei Liang, Huajun Chen, Wen Zhang

机构 * Zhejiang University(浙江大学) Ant Group(蚂蚁集团) ZJU-Ant Group Joint Lab of Knowledge Graph(浙江大学-蚂蚁集团知识图谱联合实验室)

AI总结 针对LLM在结构化知识推理中的逻辑漂移问题,提出Logits-to-Logic框架,通过logits增强和过滤模块修正输出逻辑缺陷,在多个KGQA基准上取得最佳性能。

Comments EMNLP 2026 Submission

详情

展开后加载摘要…

URL PDF HTML 收藏
2508.17320 2026-06-02 cs.LG

AdaptiveK: Complexity-Driven Sparse Autoencoders for Interpretable Language Model Representations

AdaptiveK:基于复杂度的稀疏自编码器用于可解释的语言模型表示

Yifei Yao, Hanrong Zhang, Mengnan Du

机构 * Zhejiang University(浙江大学) University of Illinois Chicago(伊利诺伊大学香槟分校) The Chinese University of Hong Kong, Shenzhen(香港中文大学(深圳))

AI总结 提出 AdaptiveK SAE,根据输入语义复杂度动态调整稀疏度,利用线性探针引导特征分配,在重构保真度、解释方差、余弦相似度和可解释性指标上优于固定稀疏度方法。

Comments Accepted by ACL 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2509.15234 2026-06-02 cs.CV

Exploring the Capabilities of Large Language Model Encoders for Image-Text Retrieval in Chest X-rays

探索大语言模型编码器在胸部X光片图像-文本检索中的能力

Hanbin Ko, Gihun Cho, Inhyeok Baek, Donguk Kim, Joonbeom Koo, Changi Kim, Dongheon Lee, Chang Min Park

机构 * Interdisciplinary Program in Bioengineering, Seoul National University Graduate School(生物工程跨学科项目,首尔国立大学研究生院) Integrated Major in Innovative Medical Science, Seoul National University Graduate School(创新医学科学整合专业,首尔国立大学研究生院) Department of Radiology, The First Affiliated Hospital, Zhejiang University School of Medicine(浙江大学医学院第一附属医院放射科) Seoul National University College of Medicine(首尔国立大学医学院) Department of Radiology, Seoul National University College of Medicine, Seoul National University Hospital(首尔国立大学医学院放射科,首尔国立大学医院) Institute of Medical and Biological Engineering, Seoul National University Medical Research Center(医学与生物工程研究所,首尔国立大学医学研究所以及) Institute of Radiation Medicine, Seoul National University Medical Research Center(放射医学研究所,首尔国立大学医学研究所以及)

AI总结 提出一种领域自适应的双向大语言模型文本编码器,通过掩码标记预测和监督对比学习训练,结合参数高效的双塔对比视觉语言框架,提升胸部X光片图像与文本的对齐和检索性能。

Comments 12 pages, 2 figures, under review

详情

展开后加载摘要…

URL PDF HTML 收藏
2508.19191 2026-06-02 cs.RO

RCM-ACT: Imitation Learning with Dynamic RCM Calibration for Autonomous Intraocular Foreign Body Removal

RCM-ACT: 基于动态RCM校准的模仿学习用于自主眼内异物取出

Yue Wang, Wenjie Deng, Haotian Xue, Di Cui, Yiqi Chen, Mingchuan Zhou, Haochao Ying, Jian Wu

机构 * College of Computer Science and Technology, Zhejiang University(浙江大学计算机科学与技术学院) State Key Laboratory of Transvascular Implantation Devices of the Second Affiliated Hospital, Zhejiang University School of Medicine(浙江大学医学院第二附属医院血管植入设备国家重点实验室) Dessight Biomedical(Dessight生物医学公司) Center for Rehabilitation Medicine, Department of Ophthalmology, Zhejiang Provincial People’s Hospital(浙江省人民医院康复医学中心、眼科部门) School of Biosystems Engineering and Food Science, Zhejiang University(浙江大学生物系统工程与食品科学学院) School of Public Health and Second Affiliated Hospital, Zhejiang University School of Medicine(浙江大学医学院公共卫生学院及第二附属医院) State Key Laboratory of Transvascular Implantation Devices of the Second Affiliated Hospital and School of Public Health, Zhejiang University School of Medicine(浙江大学医学院第二附属医院及公共卫生学院血管植入设备国家重点实验室) Zhejiang Key Laboratory of Medical Imaging Artificial Intelligence(浙江省医学影像人工智能重点实验室)

AI总结 提出RCM-ACT模仿学习框架,通过动态RCM校准和动作分块变换器解决眼内手术中的运动学不确定性,实现自主环抓取与定位,平均3D抓取偏差0.686 mm。

详情

展开后加载摘要…

URL PDF HTML 收藏
2410.02511 2026-06-02 cs.AI cs.MA

Stop Wandering, Find the Keys: LLMs Discriminate Key States for Efficient Multi-Agent Exploration

停止徘徊,找到关键:LLMs 辨别关键状态以实现高效多智能体探索

Yun Qu, Boyuan Wang, Yuhang Jiang, Jianzhun Shao, Yixiu Mao, Heming Zou, Chang Liu, Cheems Wang, Meiqin Liu, Xiangyang Ji

机构 * Department of Automation, Tsinghua University, Beijing 100084, China(清华大学自动化系) College of Electrical Engineering, Zhejiang University, Hangzhou 310027, China(浙江大学电气工程学院) National Key Laboratory of Human-Machine Hybrid Augmented Intelligence, Xi’an Jiaotong University, Xi’an 710049, China(西安交通大学人机混合增强智能国家级重点实验室)

AI总结 提出 LEMAE 方法,利用大语言模型辨别关键状态并设计子空间内在奖励和关键状态记忆树,引导多智能体高效探索,在 SMAC 和 MPE 基准上显著超越现有方法,实现 10 倍加速。

Journal ref SCIENCE CHINA Information Sciences 2026

详情

展开后加载摘要…

URL PDF HTML 收藏