arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

Northeastern University(东北大学)

2026-08-26 至 2026-08-26 共收录 7
2608.24743 2026-08-26 cs.LG cs.RO eess.SY 新提交

$(\text{DNN})^2$: Doubly Non-Negative Relaxations for Deep Neural Networks

(DNN)²:深度神经网络的双重非负松弛

Hanna Jiamei Zhang, Alan Papalia, Michael Everett, David M. Rosen

机构 * Northeastern University(东北大学) University of Michigan(密歇根大学)

AI总结 该研究针对DNN验证的松弛间隙问题,提出特征值最大化程序,使(DNN)²方法的验证边界比标准SDP更紧且可认证,为安全关键自主系统部署神经网络模块提供支撑。

Comments 6 pages, 3 figures, accepted and to be presented at 64th IEEE Conference on Decision and Control: CDC 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2608.24521 2026-08-26 cs.CL 新提交

Beyond Information Seeking: Severity-Aware Question Supervision for Proactive Medical Dialogue

超越信息获取:面向主动医疗对话的严重度感知问题监督

Chenxuan Li, Xinrong Chen, Luyan Zhang, Peidong Jia, Zhongyu Zhao, Xuecheng Shang, Peixing Wan

机构 * Peking University(北京大学) Northeastern University(东北大学) School of Basic Medical Sciences, Peking University(北京大学基础医学院)

AI总结 本研究针对主动医疗对话提出ESR严重度感知问题监督目标,在DDxPlus数据集上提升了高严重度诊断准确率、降低了漏诊率,推动主动医疗对话向感知后果的证据获取方向发展。

详情

展开后加载摘要…

URL PDF HTML 收藏
2608.23927 2026-08-26 cs.CV 新提交

GlanceWAM: Sparse Test-Time Imagination for World-Action Models

GlanceWAM:面向世界-动作模型的稀疏测试时想象方法

Linhan Wang, Zijian An, Mingyuan Zhang, Chen Dai, Yi Xu, Can Cui, Zichong Yang, Yinlin Chen, Lifeng Zhou, Chang-Tien Lu

机构 * Virginia Tech(弗吉尼亚理工大学) Drexel University(卓克索大学) Northeastern University(东北大学) Purdue University(普渡大学)

AI总结 该研究提出GlanceWAM,通过异步解耦视频DiT的想象与控制,打破世界-动作模型的速度-成功率困境,在RoboCasa、LIBERO基准测试中表现优异,推理速度达48ms/块。

详情

展开后加载摘要…

URL PDF HTML 收藏
2608.23837 2026-08-26 cs.AI 新提交

SyPS: Measuring Sycophancy Prompt Sensitivity in Large Language Models

SyPS:衡量大语言模型中的谄媚提示敏感性

Lijia Huang, Yao Fu, Sihao Ren

机构 * Northeastern University(东北大学) Case Western Reserve University(凯斯西储大学) Everpure(爱惠浦)

AI总结 本研究提出SyPS框架,通过构建控制变量的提示变体,引入SPSS指标,探究用户相关线索变化对LLMs谄媚行为的影响,发现寻求认可等线索会增加谄媚,反谄媚提示可减少谄媚。

Comments Accepted to Findings of EMNLP 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2608.23835 2026-08-26 cs.LG cs.SI 新提交

Generating Intervention Hypotheses using Explainable Explanations on Graphs: G2I, a Two-Stage Greedy Framework

基于图的可解释解释生成干预假设:G2I,一个两阶段贪心框架

Mulin Tian, Ajitesh Srivastava

机构 * University of Southern California(南加州大学) Northeastern University(东北大学)

AI总结 该研究提出两阶段贪心框架G2I,将反事实解释转化为干预设计问题,在节点层面生成可操作反事实、网络层面解决预算约束下的DNF覆盖问题,实验显示其干预策略效率优于掩码方法。

Comments 11 pages, 3 figures. Accepted at the 35th ACM International Conference on Information and Knowledge Management (CIKM 2026)

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.14180 2026-08-26 cs.CV 版本更新

Progressive Masked Refinement Self-supervised Learning for Low-Dose CT Denoising

渐进式 $\mathcal{J}$-不变自监督学习用于低剂量CT去噪

Yichao Liu, Zongru Shao, Rui Wen, Yueyang Teng, Junwen Guo

机构 * organization= IWR, Heidelberg University , city= Heidelberg , postcode= 69120 , state= Baden Württemberg , country= Germany organization= Silicon Austria Labs , city= Linz , postcode= 4040 , state= Upper Austria , country= Austria organization= Institute of Science Tokyo , addressline= , city= Tokyo , country= Japan organization= College of Medicine Biological Information Engineering, Northeastern University , city= Shenyang , postcode= 110169 , state= Liaoning , country= China organization= Key Laboratory of Intelligent Computing in Medical Image, Ministry of Education , city= Shenyang , postcode= 110169 , state= Liaoning , country= China organization= Department of Epidemiology \& Global Health, Umeå University , addressline= , city= Umeå , postcode= 90187 , country= Sweden

AI总结 提出渐进式 $\mathcal{J}$-不变学习,通过逐步盲点去噪机制和噪声注入正则化,提升低剂量CT去噪性能,在Mayo数据集上优于现有自监督方法并接近监督方法。

详情

展开后加载摘要…

URL PDF HTML 收藏
2509.13577 2026-08-26 cs.CV cs.LG cs.RO 版本更新

Adaptive Multi-Mode Out-of-Distribution Detection for Trajectory Prediction in Autonomous Vehicles

动态感知:面向自动驾驶轨迹预测的自适应多模式异常检测

Tongfei Guo, Lili Su

机构 * Department of Electrical and Computer Engineering, Northeastern University(东北大学电气与计算机工程系)

AI总结 本文提出一种自适应多模式异常检测框架,通过建模误差模式提升轨迹预测的鲁棒性,在复杂驾驶环境中实现更高效的异常检测。

Comments Accepted to the 2026 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2026)

详情

展开后加载摘要…

URL PDF HTML 收藏