arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

University of Oxford(牛津大学)

2026-07-07 至 2026-07-07 共收录 12
2607.05392 2026-07-07 cs.CV 新提交

SynCity 3000: Bootstrapping Scene-Scale 3D Diffusion

SynCity 3000:引导式构建场景级3D扩散模型

Paul Engstler, Iro Laina, Christian Rupprecht, Andrea Vedaldi

机构 * Visual Geometry Group, University of Oxford(牛津大学视觉几何组)

AI总结 针对3D场景训练数据稀缺问题,该框架将图像转3D生成器改造为卷积算子,依托新合成数据引擎微调,可生成全局连贯、支持细粒度布局控制的任意规模3D场景。

Comments Project Page: https://research.paulengstler.com/syncity-3k/

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.05230 2026-07-07 cs.LG 新提交

Video-based detection of cessation of breathing in pre-term infants using machine learning

基于机器学习的早产儿呼吸暂停视频检测方法研究

Dineo Serame, Lionel Tarassenko, Mauricio Villarroel

机构 * Institute of Biomedical Engineering, Department of Engineering Science, University of Oxford(牛津大学工程科学系生物医学工程研究所)

AI总结 针对NICU接触式呼吸监测易受干扰的问题,该研究用ResNet从视频提取胸壁运动信号检测呼吸暂停,融合常规生理信号将平衡准确率提至90.6%,提升了监测鲁棒性。

Comments Paper submitted to Computer Methods and Programs in Biomedicine (CMPB)

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.05090 2026-07-07 cs.CV 新提交

Be Indiscrete: The Benefits of Learning Continuous Spine Degeneration Severity Scores

保持离散:学习连续脊柱退变严重程度评分的益处

Maria Monzon, Andrew Zisserman, Robin Y. Park, Catherine R. Jutzeler, Amir Jamaludin

机构 * Biomedical Data Science Lab, Dept. D-HEST, ETH Zurich(生物医学数据科学实验室,D-HEST系,苏黎世联邦理工学院) Swiss Institute of Bioinformatics (SIB)(瑞士生物信息学研究所) Visual Geometry Group, Dept. of Engineering Science, University of Oxford(视觉几何组,工程科学系,牛津大学)

AI总结 研究将脊柱退变建模为连续严重程度排序问题,介绍SpineRankNet框架从腰椎MRI学习标量严重程度分数,与多类分类和序数回归对比,表明排序损失训练的模型可实现MRI扫描细粒度排序,分数能恢复分级且提升类间区分度。

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.04846 2026-07-07 cs.LG cs.AI 新提交

Pretraining Curricula Enable Selective Fine-tuning

预训练课程实现选择性微调

Sebastian A. Bruijns, Jirko Rubruck, Mia H. Whitefield, Kai J. Sandbrink, Fazl Barez, Christopher Summerfield

机构 * University of Oxford(牛津大学)

AI总结 研究预训练课程如何影响学习、泛化和微调选择性,对比平衡与不平衡预训练方式,发现不平衡预训练促进上下文学习并提高拒绝微调选择性,还扩展到合成语言学习任务。

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.04820 2026-07-07 cs.LG 新提交

KinEMbed: Decoding Kinematics from Electromyography via Cross-Modal Contrastive Learning

KinEMbed:通过跨模态对比学习从肌电图中解码运动学

Sofia Gilardini, Chenfei Ma, Kianoush Nazarpour

机构 * Department of Statistics, University of Oxford(牛津大学统计学系) School of Informatics, University of Edinburgh(爱丁堡大学信息学院)

AI总结 研究从表面肌电图解码手部运动学这一挑战,提出跨模态对比学习框架KinEMbed训练双编码器,用于手部运动学回归,在NinaPro DB8数据集上优于基线方法。

Comments ICML 2026 Workshop on Structured Data for Health, Seoul, South Korea

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.04518 2026-07-07 cs.CV 新提交

A non-invasive video-based method for individual identification of wildlife using gait dynamics

一种基于视频的非侵入性野生动物个体识别方法:利用步态动力学

Muhammad Aamir, Matthew Wijers, Sangyun Shin, Andrew Loveridge, Andrew Markham

机构 * Department of Computer Science, University of Oxford(牛津大学计算机科学系) Wildlife Conservation Research Unit, Department of Biology, University of Oxford(牛津大学生物系野生动物保护研究组)

AI总结 研究针对野生动物个体识别难题,提出基于视频的全自动方法。用深度时空表征学习,结合SAM3、ResNet18和VideoPrism模型,经微调提取特征,以余弦相似度比较步态特征实现识别,多物种实验验证了方法有效性。

Comments This article is under review in "Methods in Ecology and Evalution"

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.04293 2026-07-07 cs.CL cs.AI cs.LG stat.ML 新提交

CausalGame: Benchmarking Causal Thinking of LLM Agents in Games

因果游戏:在游戏中对大语言模型智能体的因果思维进行基准测试

Zhenhao Chen, Yongqiang Chen, Chenxi Liu, Junchi Yu, Xiangchen Song, Zijian Li, Jialin Li, Philip Torr, Bo Han, Kun Zhang

机构 * MBZUAI(穆罕默德·本·扎耶德人工智能大学) Carnegie Mellon University(卡内基梅隆大学) Hong Kong Baptist University(香港浸会大学) University of Oxford(牛津大学) New York University, Abu Dhabi(纽约大学阿布扎比分校)

AI总结 研究旨在通过因果游戏基准测试大语言模型智能体的因果思维能力,让智能体设计实验、收集数据并给出解决方案。设计含多种挑战的场景,发现当前智能体因果思维能力欠佳,为评估提供了测试平台。

Comments Zhenhao, Yongqiang, and Chenxi contributed equally to the project. A short version is accepted at the Forty-Third International Conference on Machine Learning (ICML) 2026 as an Oral presentation. Project website https://causalgame.github.io/

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.03426 2026-07-07 cs.LG cs.AI 新提交

Amortising Bayesian Experimental Design for Sequential Information Gathering in LLMs

用于大语言模型中顺序信息收集的摊销贝叶斯实验设计

Jakob Hartmann, James Harvey, Jhonathan Navott, Erik Y. Wang, Luckeciano C. Melo, Flaviu Cipcigan, Cheng Zhang, Alessandro Abate

机构 * University of Oxford(牛津大学) Ellison Institute of Technology(埃里森理工学院)

AI总结 研究大语言模型在顺序决策中信息收集问题,提出摊销顺序信息收集方法,将贝叶斯实验设计融入模型策略,经实验验证该方法能有效提升成功率并降低推理成本。

Comments 20 pages, 7 figures. Accepted to FoGen 2026: Foundations of Deep Generative Models: Understanding Memorization, Generalization, and Reasoning, an ICML 2026 workshop (non-archival)

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.02799 2026-07-07 cs.CV 新提交

Conversational Human Audio-visual Talking Dialogue Generation

对话式人类视听对话生成

Junhao Song, Lluis Guasch, Xilin He, Zhongyu Yang, Yingfang Yuan, Weicheng Xie, Linlin Shen, Haijun Lin, Shizhe Liu, Wei Pang, Siyang Song

机构 * Department of Computing, Imperial College London(伦敦帝国理工学院计算系) Department of Earth Science & Engineering, Imperial College London(伦敦帝国理工学院地球科学与工程系) Mohamed bin Zayed University of Artificial Intelligence(穆罕默德·本·扎耶德人工智能大学) Department of Computer Science, Heriot-Watt University(赫瑞瓦特大学计算机科学系) School of Computer Science, Northumbria University(诺森比亚大学计算机科学学院) College of Computer Science & Software Engineering, Shenzhen University(深圳大学计算机科学与软件学院) School of Artificial Intelligence, Shenzhen University(深圳大学人工智能学院) Guangdong Provincial Key Laboratory of Intelligent Information Processing, Shenzhen University(深圳大学广东省智能信息处理重点实验室) School of Engineering and Design, Hunan Normal University(湖南师范大学工程与设计学院) Department of Computer Science, University of Oxford(牛津大学计算机科学系) Department of Computer Science, University of Exeter(埃克塞特大学计算机科学系)

AI总结 提出CHAT框架,统一大语言模型和说话人脸模型,通过交互音频和面部行为细化模块,从单一文本提示生成多样、配对且相互响应的语音-面部对话片段,优于现有方法,合成数据集可作预训练数据。

Comments Accepted to ECCV 2026 as a main paper

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.02711 2026-07-07 cs.CV 新提交

RayTun3R: Online Camera Adaptation in 3D Foundation Models

RayTun3R:3D基础模型中的在线相机适配

Daniil Sinitsyn, Nikita Araslanov, Daniel Cremers

机构 * TU Munich(慕尼黑工业大学) Munich Center for Machine Learning(慕尼黑机器学习中心) University of Oxford(牛津大学)

AI总结 研究3D基础模型在鱼眼相机下表现不佳的问题,提出RayTun3R方法,通过学习参数高效的残差校正等,在保持预训练网络不变的情况下适配相机,减少旋转误差且参数少、无需多视图推理。

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.02692 2026-07-07 cs.CV 新提交

An Automated Multimodal Glaucoma Detection Framework Using ViT and a Stacking-Based Ensemble

使用ViT和基于堆叠的集成方法的自动多模态青光眼检测框架

Ishrat Jahan, Muhammad E. H Chowdhury, Murugappan Murugappan, Kanchon Kanti Podder, Tawsifur Rahman, Shrestha Datta, Md Sakib Abrar Hossain, Md Mosarrof Hossen, Yosra Magdi Salih Mekki, Sanjiban Sekhar Roy

机构 * Department of Computer Science and Engineering, Shahjalal University of Science and Technology(肖哈尔大学科学与技术学院计算机科学与工程系) Department of Electrical Engineering, Qatar University(卡塔尔大学电气工程系) Department of Electronics and Communication Engineering, Kuwait College of Science and Technology(科威特科学与技术学院电子与通信工程系) Department of Interdisciplinary Engineering, Kennesaw State University(肯尼斯州立大学跨学科工程系) Department of Biomedical Engineering, School of Medicine, Johns Hopkins University(约翰霍普金斯大学医学院生物医学工程系) Department of Biomedical Engineering, University of Oxford(牛津大学生物医学工程系) Department of Computer Science and Engineering, Vellore Institute of Technology(维洛雷理工学院计算机科学与工程系)

AI总结 研究在样本级和患者级设置下的青光眼检测,提出整合眼底图像与临床数据的自动多模态框架,用ViT提取特征,经典机器学习模型分类,堆叠集成优化性能,实验验证其高性能并部署为网络平台。

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.02539 2026-07-07 cs.SI cs.CL cs.LG 新提交

Beyond Satisfaction: Learning Associations Between Content, Reviews, and Well-Being

超越满意度:学习内容、评论与幸福感之间的关联

Aaron Marker, Joel Lehman, H. Andrew Schwartz

机构 * Vanderbilt University(范德比大学) University of Oxford(牛津大学) Cosmos Institute(宇宙研究所)

AI总结 研究常用满意度信号能否反映幸福感及何种内容与不同幸福感结果相关,以书籍消费为对象,发现评分等与幸福感关联松散,特定主题与不同幸福感相关,为内容推荐系统提供新目标。

详情

展开后加载摘要…

URL PDF HTML 收藏