arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

University of Cambridge(剑桥大学)

2026-03-03 至 2026-03-03 共收录 17
2603.02091 2026-03-03 cs.LG cs.AI cs.CL

Learning from Synthetic Data Improves Multi-hop Reasoning

通过合成数据学习提升多跳推理能力

Anmol Kabra, Yilun Yin, Albert Gong, Kamilė Stankevičiūtė, Dongyoung Go, Johann Lee, Katie Z. Luo, Carla P. Gomes, Kilian Q. Weinberger

机构 * Cornell University(康奈尔大学) University of Cambridge(剑桥大学) Stanford University(斯坦福大学)

AI总结 本研究通过规则生成的合成数据提升LLM多跳推理能力,发现合成数据能有效训练模型组成知识,从而在现实问答任务中表现更优。

Comments Accepted to ICLR 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.02012 2026-03-03 cs.CV cs.AI

MAP-Diff: Multi-Anchor Guided Diffusion for Progressive 3D Whole-Body Low-Dose PET Denoising

MAP-Diff: 多锚点引导的扩散模型用于渐进式三维全身低剂量PET去噪

Peiyuan Jing, Chun-Wun Cheng, Liutao Yang, Zhenxuan Zhang, Thiago V. Lima, Klaus Strobel, Antoine Leimgruber, Angelica Aviles-Rivero, Guang Yang, Javier A. Montoya-Zegarra

机构 * School of Engineering, Zurich University of Applied Sciences, CH Bioengineering Department Imperial-X, Imperial College London, UK DAMTP, University of Cambridge, UK Lucerne University Teaching Research Hospital, CH Lung Institute, Imperial College London, UK Cardiovascular Research Centre, Royal Brompton Hospital, UK School of Biomedical Engineering \& Imaging Sciences, King's College London, UK Yau Mathematical Sciences Center, Tsinghua University, CN

AI总结 MAP-Diff通过多锚点引导的扩散模型实现低剂量PET图像的渐进式去噪,提升PSNR和SSIM,降低NMAE,优于多种基线方法。

Comments 8 pages, 3 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.01990 2026-03-03 cs.AI cs.CL cs.CV

According to Me: Long-Term Personalized Referential Memory QA

根据我:长期个性化参照记忆问答

Jingbiao Mei, Jinghong Chen, Guangyu Yang, Xinyu Hou, Margaret Li, Bill Byrne

机构 * Department of Engineering, University of Cambridge, United Kingdom(剑桥大学工程系) Department of Physics, University of Cambridge, United Kingdom(剑桥大学物理系) Independent Researcher(独立研究者)

AI总结 本文提出ATM-Bench,首个多模态多来源个性化参照记忆问答基准,并提出Schema-Guided Memory方法提升记忆推理性能

Comments Preprint

详情

展开后加载摘要…

URL PDF HTML 收藏
2407.08086 2026-03-03 cs.LG stat.CO stat.ML

The GeometricKernels Package: Heat and Matérn Kernels for Geometric Learning on Manifolds, Meshes, and Graphs

几何核包:用于流形、网格和图上几何学习的热核和Matérn核

Peter Mostowsky, Vincent Dutordoir, Iskander Azangulov, Noémie Jaquier, Michael John Hutchinson, Aditya Ravuri, Leonel Rozo, Alexander Terenin, Viacheslav Borovitskiy

机构 * University of Cambridge(剑桥大学) University of Oxford(牛津大学) KTH Royal Institute of Technology(皇家理工学院) Italian Institute of Artificial Intelligence for Industry(意大利人工智能工业研究所) Cornell University(康奈尔大学) ETH Zürich and University of Edinburgh(苏黎世联邦理工学院和爱丁堡大学)

AI总结 GeometricKernels 是一个用于几何学习的Python包,实现了热核和Matérn核,支持在流形、网格和图上进行不确定性量化和自动微分。

Journal ref Journal of Machine Learning Research, 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.01414 2026-03-03 cs.RO

Jailbreaking Embodied LLMs via Action-level Manipulation

通过动作层面操控实现体感大语言模型的突破

Xinyu Huang, Qiang Yang, Leming Shen, Zijing Ma, Yuanqing Zheng

机构 * The Hong Kong Polytechnic University, Hong Kong SAR, China(香港理工大学) University of Cambridge, Cambridge, United Kingdom(剑桥大学)

AI总结 Blindfold通过动作层面操控实现对体感大语言模型的突破,其攻击成功率显著高于现有基线,凸显了对后果感知防御机制的迫切需求。

Comments This paper has been officially accepted for ACM SenSys 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.18182 2026-03-03 cs.LG cs.AI

Capabilities Ain't All You Need: Measuring Propensities in AI

能力并非全部所需:测量AI倾向性

Daniel Romero-Alvarado, Fernando Martínez-Plumed, Lorenzo Pacchiardi, Hugo Save, Siddhesh Milind Pawar, Behzad Mehrbakhsh, Pablo Antonio Moreno Casares, Ben Slater, Paolo Bova, Peter Romero, Zachary R. Tidler, Jonathan Prunty, Luning Sun, Jose Hernandez-Orallo

机构 * Valencian Research Institute of Artificial Intelligence, Universitat Politècnica de València, Valencia, Spain University of Copenhagen, Denmark work done while at University of Cambridge Existential Risk Observatory, Amsterdam, Netherlands Leverhulme Centre for the Future of Intelligence, University of Cambridge The Psychometrics Centre, University of Cambridge Department of Computing \& Games, University of Teesside Georgia Institute of Technology University of Cambridge

AI总结 本文提出首个测量AI倾向性的正式框架,通过双逻辑模型评估模型倾向性对任务性能的影响,并展示结合倾向性和能力可提升预测效果。

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.08630 2026-03-03 cs.CL

ExPO-HM: Learning to Explain-then-Detect for Hateful Meme Detection

ExPO-HM:为仇恨表情包检测学习解释-然后检测

Jingbiao Mei, Mingsheng Sun, Jinghong Chen, Pengda Qin, Yuhong Li, Da Chen, Bill Byrne

机构 * Department of Engineering, University of Cambridge(剑桥大学工程系) Xiaohongshu Inc.(小红书公司) University of Bath(巴斯大学) Tencent Company, China(腾讯公司) Alibaba Group(阿里巴巴集团)

AI总结 ExPO-HM通过结合SFT预热、GRPO与课程学习及CDE,提升仇恨表情包检测的解释性和准确性。

Comments ICLR 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2506.09434 2026-03-03 cs.MA cs.AI cs.LG

When Is Diversity Rewarded in Cooperative Multi-Agent Learning?

在合作多智能体学习中,多样性何时会被奖励?

Michael Amir, Matteo Bettini, Amanda Prorok

机构 * Department of Computer Science and Technology University of Cambridge(计算机科学与技术系剑桥大学)

AI总结 研究通过理论分析和算法验证,探讨了多智能体学习中异质性团队奖励机制,提出HetGPS算法用于发现异质性优势的奖励场景。

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.13061 2026-03-03 cs.CL cs.AI cs.CV cs.LG

Robust Adaptation of Large Multimodal Models for Retrieval Augmented Hateful Meme Detection

鲁棒适应大规模多模态模型用于检索增强的仇恨表情包检测

Jingbiao Mei, Jinghong Chen, Guangyu Yang, Weizhe Lin, Bill Byrne

机构 * Department of Engineering University of Cambridge(工程系剑桥大学)

AI总结 本文提出了一种鲁棒适应框架,用于提升大规模多模态模型在检索增强的仇恨表情包检测中的性能与泛化能力,同时提高模型的可解释性。

Comments EMNLP 2025 Main (Oral)

详情

展开后加载摘要…

URL PDF HTML 收藏
2410.10045 2026-03-03 cs.RO cs.AI cs.LG

Neuro-Symbolic Skill Discovery for Conditional Multi-Level Planning

神经符号技能发现用于条件多级规划

Hakan Aktas, Yigit Yildirim, Ahmet Firat Gamsiz, Deniz Bilge Akkoc, Erhan Oztop, Emre Ugur

机构 * Department of Computer Science and Technology The University of Cambridge(计算机科学与技术系 剑桥大学) Department of Computer Engineering Bogazici University(计算机工程系 boazici大学) SISREC Osaka University(Osaka大学SISREC)

AI总结 本文提出了一种神经符号学习架构,通过少量演示获取可推广的高级符号技能,并在多级规划中实现复杂任务的执行。

Comments 18 pages, 4 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2406.16227 2026-03-03 stat.ML cs.LG stat.ME

VICatMix: variational Bayesian clustering and variable selection for discrete biomedical data

VICatMix:变分贝叶斯聚类与变量选择用于离散生物医学数据

Jackie Rao, Paul D. W. Kirk

机构 * MRC Biostatistics Unit University of Cambridge East Forvie Building(医学研究理事会生物统计学单位剑桥大学东部大楼) MRC Biostatistics Unit and CRUK Cambridge Centre Ovarian Programme and Cambridge Institute of Therapeutic Immunology and Infectious Disease (CITIID) University of Cambridge(医学研究理事会生物统计学单位和癌症研究英国委员会剑桥中心卵巢计划及剑桥治疗免疫学和传染病研究所(CITIID)剑桥大学)

AI总结 VICatMix通过变分贝叶斯方法实现高效聚类和变量选择,适用于高维生物医学数据,提升癌症亚型分组和驱动基因发现的准确性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2508.16479 2026-03-03 eess.IV cs.AI cs.CV

Disentangled Multi-modal Learning of Histology and Transcriptomics for Cancer Characterization

解耦的多模态学习:组织学与转录组学用于癌症表征

Yupei Zhang, Xiaofei Wang, Anran Liu, Lequan Yu, Chao Li

机构 * Department of Clinical Neurosciences, University of Cambridge, UK(剑桥大学临床神经科学系) Department of Health Technology & Informatics, The Hong Kong Polytechnic University(香港理工大学健康科技与信息学系) Department of Statistics and Actuarial Science, The University of Hong Kong(香港大学统计与精算科学系) Department of Clinical Neurosciences and Department of Applied Mathematics and Theoretical Physics, University of Cambridge(剑桥大学临床神经科学系和应用数学与理论物理系;邓迪大学科学与工程学院和医学院) School of Science and Engineering and School of Medicine, University of Dundee, UK

AI总结 本文提出了解耦的多模态学习框架,通过分解组织学和转录组数据以提高癌症表征的准确性和实用性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.19949 2026-03-03 cs.LG eess.SP

Machine-learning for photoplethysmography analysis: Benchmarking feature, image, and signal-based approaches

机器学习在光体积脉搏波形成像分析中的应用:特征、图像和信号基于方法的基准测试

Mohammad Moulaeifard, Loic Coquelin, Mantas Rinkevičius, Andrius Sološenko, Oskar Pfeffer, Ciaran Bench, Nando Hegemann, Sara Vardanega, Manasi Nandi, Jordi Alastruey, Christian Heiss, Vaidotas Marozas, Andrew Thompson, Philip J. Aston, Peter H. Charlton, Nils Strodthoff

机构 * Carl von Ossietzky Universität Oldenburg(奥尔登堡卡尔·冯·奥西特齐大学) Laboratoire national de métrologie et d’essais(国家计量与测试实验室) Biomedical Engineering Institute, Kaunas University of Technology(克莱萨斯理工大学生物医学工程研究所) Physikalisch-Technische Bundesanstalt(德国物理技术部) National Physical Laboratory(国家物理实验室) King’s College London(伦敦国王学院) University of Surrey(萨里大学) Department of Public Health and Primary Care, University of Cambridge(剑桥大学公共卫生与初级保健系)

AI总结 本文通过对比特征、图像和信号三种方法,发现深度神经网络在PPG数据分析中表现最佳,尤其在血压预测和心房颤动预测任务中。

Comments 39 pages, 9 figures, code available at https://gitlab.com/qumphy/d1-code

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.19167 2026-03-03 cs.LG eess.SP

Generalizable deep learning for photoplethysmography-based blood pressure estimation -- A Benchmarking Study

可泛化深度学习用于基于光体积脉搏波的血压估计——一项基准研究

Mohammad Moulaeifard, Peter H. Charlton, Nils Strodthoff

机构 * Carl von Ossietzky Universität Oldenburg(奥尔登堡卡尔·冯·奥西特齐大学) University of Cambridge(剑桥大学)

AI总结 本文提出了一种基于样本域适应的深度学习方法,用于改进PPG血压估计在不同数据集间的泛化能力,并通过实验验证了分布外性能的挑战。

Comments 20 pages, 5 figures, code available at https://github.com/AI4HealthUOL/ppg-ood-generalization

Journal ref Machine Learning: Health 1(1):010501, 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.12063 2026-03-03 stat.ML cs.LG math.OC math.ST stat.ME stat.TH

Low-Rank Thinning

低秩稀疏化

Annabelle Michael Carrell, Albert Gong, Abhishek Shetty, Raaz Dwivedi, Lester Mackey

机构 * University of Cambridge(剑桥大学) Microsoft Research New England(微软研究院新英格兰分部) MIT(麻省理工学院)

AI总结 本文提出了一种低秩分析的子高斯稀疏化方法,适用于任何分布和核,能够保证高质量压缩,并在变换器注意力近似、随机梯度训练加速和分布区分中取得改进。

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.00102 2026-03-03 cs.RO

Designing Social Robots with Ethical, User-Adaptive Explainability in the Era of Foundation Models

在基础模型时代设计具有伦理性和用户适应性的社交机器人

Fethiye Irmak Dogan, Alva Markelius, Hatice Gunes

机构 * University of Cambridge(剑桥大学)

AI总结 本文提出在基础模型驱动的社交机器人中,需通过伦理性和用户适应性的可解释性设计,解决适应与解释委托给基础模型带来的挑战。

Comments Companion Proceedings of the 21st ACM/IEEE International Conference on Human-Robot Interaction

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.00063 2026-03-03 cs.CY cs.AI cs.LG

Measuring What AI Systems Might Do: Towards A Measurement Science in AI

衡量人工智能系统可能的行为:迈向人工智能中的测量科学

Konstantinos Voudouris, Mirko Thalmann, Alex Kipnis, José Hernández-Orallo, Eric Schulz

机构 * Institute for Human-Centered AI(以人为本的人工智能研究所) Leverhulme Centre for the Future of Intelligence(未来智能研究中心) University of Cambridge(剑桥大学)

AI总结 本文提出了一种基于倾向属性的AI评估方法,强调通过因果相关性、操作化测量和经验映射来衡量AI系统的稳定特征。

详情

展开后加载摘要…

URL PDF HTML 收藏