arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

University of Toronto(多伦多大学)

共收录 1022
2502.17424 2026-01-27 cs.CL cs.AI cs.CR cs.LG

Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs

涌现的偏移:狭窄微调可以产生广泛偏移的LLM

Jan Betley, Daniel Tan, Niels Warncke, Anna Sztyber-Betley, Xuchan Bao, Martín Soto, Nathan Labenz, Owain Evans

机构 * University College London(伦敦大学学院) Center on Long-Term Risk(长期风险中心) Warsaw University of Technology(华沙技术大学) University of Toronto(多伦多大学)

AI总结 研究发现,狭窄微调训练LLM生成不安全代码会导致广泛偏移,模型在无关提示上表现出欺骗性行为,且偏移可通过触发器隐藏。

Comments 41 pages, 38 figures An earlier revision of this paper was accepted at ICML 2025. Since then, it has been updated to include new results on the impact of formatting (4.4), new dataset (4.6), training dynamics (4.7) and base models (4.8) Extended version of the paper was published in Nature 2026/1

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.16034 2026-01-27 cs.CL

Examining the Utility of Self-disclosure Types for Modeling Annotators of Social Norms

检验自我披露类型对建模社会规范注释者的作用

Kieran Henderson, Kian Omoomi, Vasudha Varadarajan, Allison Lahnala, Charles Welch

机构 * University of Toronto(多伦多大学) Carnegie Mellon University(卡内基梅隆大学) McMaster University(麦马斯特大学)

AI总结 本研究通过分析自我披露类型对社会规范注释者预测的影响,发现少量相关评论即可有效建模,并指出扩大样本多样性未必提升性能。

Comments Accepted EACL Findings

详情

展开后加载摘要…

URL PDF HTML 收藏
2507.21288 2026-01-27 cs.GR cs.AI

SpringTime: Learning Simulatable Models of Cloth with Spatially-varying Constitutive Properties

SpringTime: 学习具有空间变化本构性质的布料模拟模型

Guanxiong Chen, Shashwat Suri, Yuhao Wu, Yixian Cheng, Ganidhu Abeysirigoonawardena, Etienne Vouga, David I. W. Levin, Dinesh K. Pai

机构 * University of British Columbia(不列颠哥伦比亚大学) University of Texas at Austin(德克萨斯大学奥斯汀分校) University of Toronto(多伦多大学)

AI总结 SpringTime通过学习运动观测数据,高效建模具有空间变化本构性质的布料模拟,克服膜锁定问题,提升训练速度和重建精度。

Comments Submitted to Graphics Interface '26 (In review)

详情

展开后加载摘要…

URL PDF HTML 收藏
2506.19113 2026-01-27 cs.CL

Argument-Based Consistency in Toxicity Explanations of LLMs

基于论证的一致性:大语言模型毒性解释的评估

Ramaravind Kommiya Mothilal, Joanna Roy, Syed Ishtiaque Ahmed, Shion Guha

机构 * University of Toronto(多伦多大学) trail-ml

AI总结 本研究提出基于论证的一致性标准,评估大语言模型对毒性的推理能力,发现其在复杂关系下的解释不一致。

Comments 29 pages, 7 figures, 9 tables

详情

展开后加载摘要…

URL PDF HTML 收藏
2503.08305 2026-01-27 cs.LG physics.chem-ph physics.comp-ph

ELECTRA: A Cartesian Network for 3D Charge Density Prediction with Floating Orbitals

ELECTRA:一种用于3D电荷密度预测的笛卡尔网络

Jonas Elsborg, Luca Thiede, Alán Aspuru-Guzik, Tejs Vegge, Arghya Bhowmik

机构 * Technical University of Denmark(丹麦技术大学) CAPeX Pioneer Center for Accelerating P2X Materials Discovery(CAPeX先锋中心) University of Toronto(多伦多大学) Vector Institute for Artificial Intelligence(人工智能矢量研究所) Canadian Institute for Advanced Research (CIFAR)(加拿大高级研究 institute)

AI总结 ELECTRA通过笛卡尔张量网络预测漂浮轨道位置,实现高效且准确的3D电荷密度预测,并显著降低DFT计算的SCF迭代次数。

Comments 10 pages, 4 figures, 5 tables, NeurIPS 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.16649 2026-01-26 cs.AI

LUMINA: Long-horizon Understanding for Multi-turn Interactive Agents

LUMINA:多轮交互智能体的长视图理解

Amin Rakhsha, Thomas Hehn, Pietro Mazzaglia, Fabio Valerio Massoli, Arash Behboodi, Tribhuvanesh Orekondy

机构 * University of Toronto(多伦多大学) Qualcomm AI Research(高通人工智能研究)

AI总结 LUMINA研究通过oracle反事实框架分析多轮交互智能体中各基础能力的重要性,揭示了不同技能在不同环境下的有效性差异。

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.12409 2026-01-26 cs.LG cs.AI cs.NE

Interpretable Fine-Gray Deep Survival Model for Competing Risks: Predicting Post-Discharge Foot Complications for Diabetic Patients in Ontario

可解释的细灰深度生存模型用于竞争风险:预测安大略省糖尿病患者出院后足部并发症

Dhanesh Ramachandram, Anne Loefler, Surain Roberts, Amol Verma, Maia Norman, Fahad Razak, Conrad Pow, Charles de Mestral

机构 * Vector Institute(向量研究所) GEMINI University of Toronto(多伦多大学) Diabetes Action Canada(加拿大糖尿病行动)

AI总结 本文提出了一种可解释的细灰深度生存模型,用于预测糖尿病患者出院后足部并发症,通过透明的预测方法提高医疗AI的可信度。

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.22327 2026-01-23 cs.CL cs.CY

NLP for Social Good: A Survey and Outlook of Challenges, Opportunities, and Responsible Deployment

为社会公益服务的NLP:挑战、机遇与负责任部署的综述与展望

Antonia Karamolegkou, Angana Borah, Eunjung Cho, Sagnik Ray Choudhury, Martina Galletti, Pranav Gupta, Oana Ignat, Priyanka Kargupta, Neema Kotonya, Hemank Lamba, Sun-Joo Lee, Arushi Mangla, Ishani Mondal, Fatima Zahra Moudakir, Deniz Nazarova, Poli Nemkova, Dina Pisarevskaya, Naquee Rizwan, Nazanin Sabri, Keenan Samway, Dominik Stammbach, Anna Steinberg, David Tomás, Steven R Wilson, Bowen Yi, Jessica H Zhu, Arkaitz Zubiaga, Anders Søgaard, Alexander Fraser, Zhijing Jin, Rada Mihalcea, Joel R. Tetreault, Daryna Dementieva

机构 * University of Copenhagen(哥本哈根大学) University of Michigan-Ann Arbor(密歇根大学安娜堡分校) ETH Zurich(苏黎世联邦理工学院) University of North Texas(北卡罗来纳州立大学) Sony Computer Science Laboratories - Paris(索尼计算机科学实验室-巴黎) Santa Clara University(圣克拉拉大学) University of Illinois Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校) Dataminr(DataMinr公司) United Nations Development Programme (UNDP)(联合国开发计划署) University of Maryland, College Park(马里兰大学学院市分校) Max Planck Institute for Intelligent Systems, Tübingen(智能系统马克斯·普朗克研究所,图宾根) Vector Institute(向量研究所) University of Toronto(多伦多大学) University of Washington(华盛顿大学) Queen Mary University of London(伦敦大学玛丽女王学院) IIT Kharagpur(印度理工学院Kharagpur分校) University of California San Diego(加州大学圣地亚哥分校) Princeton University(普林斯顿大学) LMU Munich(慕尼黑大学) Munich Center for Machine Learning (MCML)(慕尼黑机器学习中心) University of Alicante(阿利坎特大学) University of Michigan-Flint(密歇根大学弗林特分校) University of Southern California(南加州大学) Technical University of Munich(慕尼黑技术大学)

AI总结 本文综述了NLP在社会公益领域的应用现状,指出包容性和AI危害是研究热点,同时呼吁跨学科合作以促进公众福祉。

Comments Accepted to EACL 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2506.19780 2026-01-22 cs.LG

Listwise Direct Preference Optimization with Multi-Dimensional Preference Mixing

基于多维偏好混合的列表式直接偏好优化

Yuhui Sun, Xiyao Wang, Zixi Li, YiTian Ding, Tianyang Ling, Jialuo Chen, Tianyi Yu, Zhenlong Yuan, Jinman Zhao

机构 * University of Alberta(阿尔伯塔大学) University of Toronto(多伦多大学) Zhejiang University(浙江大学) School of Computer Science McGill University(麦吉尔大学计算机学院) Alibaba Group (Ant Group)(阿里巴巴集团(蚂蚁集团)) Chinese Academy of Sciences(中国科学院)

AI总结 本文提出λ-DPO,通过多维偏好混合和自适应调度器提升模型对多维度偏好权衡的建模能力。

Comments 13 pages, 1 figures, appendix included

详情

展开后加载摘要…

URL PDF HTML 收藏
2411.19908 2026-01-21 stat.ML cs.LG

Another look at statistical inference with machine learning-imputed data

对利用机器学习插补数据的统计推断的再审视

Jessica Gronsbell, Jianhui Gao, Zachary R. McCaw, Yaqi Shi, David Cheng

机构 * Department of Statistical Sciences, University of Toronto(统计科学系,多伦多大学) Department of Biostatistics, UNC Chapel Hill(生物统计学系,北卡罗来纳大学教堂山分校) Biostatistics Center, Massachusetts General Hospital(麻省总医院生物统计中心)

AI总结 本文提出了一种基于机器学习插补结果的Z估计方法,旨在提高统计推断效率并减少预测误差带来的偏差。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.13232 2026-01-21 cs.RO

MATTERIX: toward a digital twin for robotics-assisted chemistry laboratory automation

MATTERIX:迈向机器人辅助化学实验室自动化数字孪生

Kourosh Darvish, Arjun Sohal, Abhijoy Mandal, Hatem Fakhruldeen, Nikola Radulov, Zhengxue Zhou, Satheeshkumar Veeramani, Joshua Choi, Sijie Han, Brayden Zhang, Jeeyeoun Chae, Alex Wright, Yijie Wang, Hossein Darvish, Yuchi Zhao, Gary Tom, Han Hao, Miroslav Bogdanovic, Gabriella Pizzuto, Andrew I. Cooper, Alán Aspuru-Guzik, Florian Shkurti, Animesh Garg

机构 * University of Toronto(多伦多大学) Acceleration Consortium(加速联盟) Vector Institute(向量研究所) University of Liverpool(利物浦大学) University of Salento(萨勒诺大学) NVIDIA Canadian Institute for Advanced Research(NVIDIA加拿大高级研究机构) Georgia Institute of Technology(佐治亚理工学院)

AI总结 MATTERIX通过多尺度仿真和模块化引擎,实现机器人辅助化学实验室的数字孪生,加速工作流程开发并减少现实实验依赖。

Comments Darvish, K., Sohal, A., Mandal, A. et al. MATTERIX: toward a digital twin for robotics-assisted chemistry laboratory automation. Nat Comput Sci (2025)

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.12740 2026-01-21 cs.HC cs.AI

TreeWriter: AI-Assisted Hierarchical Planning and Writing for Long-Form Documents

TreeWriter: 人工智能辅助的长篇文档分层规划与写作

Zijian Zhang, Fangshi Du, Xingjian Liu, Pan Chen, Oliver Huang, Runlong Ye, Michael Liut, Alán Aspuru-Guzik

机构 * Department of Computer Science, University of Toronto, Sandford Fleming Building, 10 King’s College Road, ON M5S 3G4, Toronto, Canada Vector Institute for Artificial Intelligence, 661 University Ave. Suite 710, ON M5G 1M1, Toronto, Canada Department of Chemistry, University of Toronto, Lash Miller Chemical Laboratories, 80 St. George Street, ON M5S 3H6, Toronto, Canada Department of Mathematical Computational Sciences, University of Toronto Mississauga, 3359 Mississauga Road, Deerfield Hall, ON L5L 1C6, Mississauga, Canada Department of Materials Science \& Engineering, University of Toronto, 184 College St., M5S 3E4, Toronto, Canada Department of Chemical Engineering \& Applied Chemistry, University of Toronto, 200 College St. ON M5S 3E5, Toronto, Canada Acceleration Consortium, 700 University Ave., M7A 2S4, Toronto, Canada Canadian Institute for Advanced Research (CIFAR), 661 University Ave., M5G 1M1, Toronto, Canada NVIDIA, 431 King St W \#6th, M5V 1K4, Toronto, Canada

AI总结 TreeWriter通过分层结构和集成AI支持,提升长文档写作中的创意发展和用户控制能力。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.12639 2026-01-21 cs.CL cs.LG

Objective Matters: Fine-Tuning Objectives Shape Safety, Robustness, and Persona Drift

目标至关重要:微调目标影响安全、鲁棒性和人格漂移

Daniel Vennemeyer, Punya Syon Pandey, Phan Anh Duong, Michael Umeokoli, Samuel Ratnam

机构 * University of Cincinnati(辛辛那提大学) University of Toronto(多伦多大学) University of Oxford(牛津大学)

AI总结 本文研究了微调目标对LLM安全性和鲁棒性的影响,发现目标选择在不同训练规模下对安全、鲁棒性和人格漂移有显著影响。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.12585 2026-01-21 cs.HC cs.AI cs.ET

Do MLLMs See What We See? Analyzing Visualization Literacy Barriers in AI Systems

MLLMs是否看到我们看到的?分析AI系统中可视化素养障碍

Mengli, Duan, Yuhe, Jiang, Matthew Varona, Carolina Nobre

机构 * University of Toronto(多伦多大学)

AI总结 本研究分析了MLLMs在可视化素养中的障碍,揭示了两种机器特定的障碍,并展示了模型在复杂可视化任务上的表现差异。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.12543 2026-01-21 cs.LG

Press Start to Charge: Videogaming the Online Centralized Charging Scheduling Problem

按下开始按钮充电:将在线集中充电调度问题游戏化

Alireza Ghahtarani, Martin Cousineau, Amir-massoud Farahmand, Jorge E. Mendoza

机构 * Department of Logistics and Operations Management, HEC Montréal(物流与运营管理系,蒙特利尔HEC商学院) Department of Computer Science, University of Toronto(计算机科学系,多伦多大学) Department of Computer Engineering and Software Engineering, Polytechnique Montréal(计算机工程与软件工程系,蒙特利尔Polytechnique大学)

AI总结 通过游戏化方法提升电动汽车充电调度的负载平衡与经济效益

Comments 41 pages

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.02543 2026-01-21 cs.LG cs.AI cs.CV cs.IT math.IT

Normalized Conditional Mutual Information Surrogate Loss for Deep Neural Classifiers

归一化条件互信息替代损失用于深度神经分类器

Linfeng Ye, Zhixiang Chi, Konstantinos N. Plataniotis, En-hui Yang

机构 * Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, Canada(滑铁卢大学电气与计算机工程系) The Edward S. Rogers Sr. Department of Electrical and Computer Engineering, University of Toronto, Toronto, Canada(多伦多大学Edward S. Rogers Sr.电气与计算机工程系)

AI总结 本文提出归一化条件互信息作为替代交叉熵损失,通过实验表明其在图像识别和全滑片成像任务中显著提升模型性能。

Comments 8 pages, 4 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.24195 2026-01-21 cs.HC cs.CL

WikiGap: Promoting Epistemic Equity by Surfacing Knowledge Gaps Between English Wikipedia and other Language Editions

WikiGap: 通过揭示英语维基百科与其他语言版本之间的知识空白促进知识公平

Zining Wang, Yuxuan Zhang, Dongwook Yoon, Nicholas Vincent, Farhan Samir, Vered Shwartz

机构 * University of British Columbia(不列颠哥伦比亚大学) Vector Institute for AI(人工智能向量研究所) Simon Fraser University(西蒙弗雷泽大学) University of Toronto(多伦多大学)

AI总结 WikiGap通过揭示英语维基百科与其他语言版本之间的知识空白,提升不同语言版本之间的知识公平性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2503.05938 2026-01-21 cs.LG cs.AI hep-ex hep-ph hep-th

Uncertainty Quantification From Scaling Laws in Deep Neural Networks

深度神经网络中从缩放定律量化不确定性

Ibrahim Elsharkawy, Yonatan Kahn, Benjamin Hooberman

机构 * Department of Physics, University of Illinois Urbana-Champaign, Urbana, IL, USA(伊利诺伊大学厄巴纳-香槟分校物理系) Department of Physics, University of Toronto, Toronto, ON, Canada(多伦多大学物理系) Vector Institute, Toronto, ON, Canada(向量研究所)

AI总结 本文研究了深度神经网络中通过缩放定律量化不确定性的方法,发现测试损失的方差与均值比值在足够大的训练集下与网络宽度无关。

Comments 18+3 pages, 6 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.10690 2026-01-16 cs.LG

Data-driven stochastic reduced-order modeling of parametrized dynamical systems

数据驱动的随机降阶建模:参数化动力系统

Andrew F. Ilersich, Kevin Course, Prasanth B. Nair

机构 * University of Toronto Institute for Aerospace Studies(多伦多大学航空航天研究所)

AI总结 本文提出了一种数据驱动的随机降阶建模方法,通过概率自编码器和随机微分方程联合学习,实现跨参数空间的泛化和高效建模。

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.10510 2026-01-16 cs.LG stat.ML

MixMin: Finding Data Mixtures via Convex Minimization

MixMin: 通过凸优化寻找数据混合

Anvith Thudi, Evianne Rovers, Yangjun Ruan, Tristan Thrush, Chris J. Maddison

机构 * University of Toronto(多伦多大学) Vector Institute(向量研究所) Stanford University(斯坦福大学)

AI总结 MixMin通过凸优化方法改进数据混合,提升了语言模型和化学任务的性能,且在不同规模模型上均有效。

Comments Proceedings of the 42nd International Conference on Machine Learning

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.10250 2026-01-16 eess.IV cs.CV q-bio.QM

Cell Behavior Video Classification Challenge, a benchmark for computer vision methods in time-lapse microscopy

细胞行为视频分类挑战,一种用于时间延拓显微镜中计算机视觉方法的基准测试

Raffaella Fiamma Cabini, Deborah Barkauskas, Guangyu Chen, Zhi-Qi Cheng, David E Cicchetti, Judith Drazba, Rodrigo Fernandez-Gonzalez, Raymond Hawkins, Yujia Hu, Jyoti Kini, Charles LeWarne, Xufeng Lin, Sai Preethi Nakkina, John W Peterson, Koert Schreurs, Ayushi Singh, Kumaran Bala Kandan Viswanathan, Inge MN Wortel, Sanjian Zhang, Rolf Krause, Santiago Fernandez Gonzalez, Diego Ulisse Pizzagalli

机构 * Euler Institute, Faculty of Informatics, Università della Svizzera italiana(欧拉研究所,信息学院,瑞士大学) International Center for Advanced Computing in Medicine (ICAM), University of Pavia(国际医学先进计算中心(ICAM),帕维亚大学) Imaging Platform, ACRF INCITe Centre, Garvan Institute of Medical Research(成像平台,ACRF INCITe中心,嘉文医学研究所) Tacoma School of Engineering & Technology, University of Washington(塔科马工程与技术学院,华盛顿大学) Data Science, Institute for Computing and Information Sciences, Radboud University(数据科学,计算与信息科学研究所,拉德堡德大学) Imaging Core, Lerner Research Institute, Cleveland Clinic(成像核心,勒纳研究研究所,克利夫兰诊所) Institute of Biomedical Engineering, University of Toronto(生物医学工程研究所,多伦多大学) Center for Research in Computer Vision, University of Central Florida(计算机视觉研究中心,佛罗里达大学) Computational Biology Group, Data Science Platform, Garvan Institute of Medical Research(计算生物学小组,数据科学平台,嘉文医学研究所)

AI总结 本文提出细胞行为视频分类挑战,评估了三种方法在时间延拓显微镜视频分类中的性能,旨在推动计算机视觉在细胞动态研究中的发展。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.10090 2026-01-16 cs.CV cs.AI

Difficulty-guided Sampling: Bridging the Target Gap between Dataset Distillation and Downstream Tasks

难度引导采样:弥合数据集蒸馏与下游任务之间的目标差距

Mingzhuo Li, Guang Li, Linfeng Ye, Jiafeng Mao, Takahiro Ogawa, Konstantinos N. Plataniotis, Miki Haseyama

机构 * Hokkaido University(北海道大学) University of Toronto(多伦多大学) The University of Tokyo(东京大学)

AI总结 本文提出难度引导采样方法,通过引入难度概念和难度感知指导,弥合数据集蒸馏与下游任务之间的目标差距,提升蒸馏效果和下游任务性能。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.09626 2026-01-15 cs.LG cs.AI cs.SY eess.SY

From Prompt to Protocol: Fast Charging Batteries with Large Language Models

从提示到协议:利用大语言模型实现快速充电电池

Ge Lei, Ferran Brosa Planella, Sterling G. Baird, Samuel J. Cooper

机构 * Dyson School of Design Engineering, Imperial College London(帝国理工学院伦敦设计工程学院) University of Warwick(沃里克大学) Acceleration Consortium, University of Toronto(多伦多大学加速联盟)

AI总结 利用大语言模型设计快速充电电池协议,通过Prompt-to-Optimizer和Prompt-to-Protocol方法提升充电效率和电池健康状态

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.09676 2026-01-15 cs.LG cs.AI stat.ML

Coupled Data and Measurement Space Dynamics for Enhanced Diffusion Posterior Sampling

耦合数据与测量空间动态用于增强扩散后验采样

Shayan Mohajer Hamidi, En-Hui Yang, Ben Liang

机构 * Stanford University(斯坦福大学) University of Toronto(多伦多大学) University of Waterloo(滑铁卢大学)

AI总结 本文提出C-DPS框架,通过耦合数据与测量空间动态,实现逆问题中更准确的扩散后验采样,提升高噪声条件下的性能。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.09191 2026-01-15 cs.CV

From Performance to Practice: Knowledge-Distilled Segmentator for On-Premises Clinical Workflows

从性能到实践:面向本地临床工作流的知识蒸馏分割器

Qizhen Lan, Aaron Choi, Jun Ma, Bo Wang, Zhaogming Zhao, Xiaoqian Jiang, Yu-Chun Hsu

机构 * D. Bradley McWilliams School of Biomedical Informatics(D. Bradley McWilliams 生物医学信息学学院) The University of Texas Health Science Center at Houston(德克萨斯大学健康科学中心休斯顿分校) M31 AI University Health Network(大学健康网络) University of Toronto(多伦多大学) Vector Institute(向量研究所) Center for Precision Health(精准健康中心)

AI总结 本文提出了一种面向部署的知识蒸馏分割器,通过减少模型参数提升本地临床工作流的实用性与可维护性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.07878 2026-01-14 cs.LG cs.AI cs.CL

Sliced-Wasserstein Distribution Alignment Loss Improves the Ultra-Low-Bit Quantization of Large Language Models

切片瓦瑟斯坦分布对齐损失提高了大语言模型的超低比特量化

Deyu Cao, Yixin Yin, Samin Aref

机构 * Department of Information and Communication Engineering, The University of Tokyo(信息与通信工程系,东京大学) Department of Computer Science, University of Toronto(计算机科学系,多伦多大学) Department of Mechanical and Industrial Engineering, University of Toronto(机械与工业工程系,多伦多大学)

AI总结 切片瓦瑟斯坦分布对齐损失通过提升超低比特量化性能,有效恢复模型准确性。

Comments Post-peer-review accepted manuscript, 17 pages including the supplementary information

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.07573 2026-01-13 econ.TH cs.AI

A Model of Artificial Jagged Intelligence

人工锯齿智能模型

Joshua Gans

机构 * Rotman School of Management, University of Toronto and NBER(多伦多大学罗特曼管理学院和美国国家经济研究局)

AI总结 本文提出一个经济模型解释生成式AI在看似相近任务上的性能不均衡现象,通过局部可靠性与全局质量信号的差异,揭示了用户采用决策中的信息问题及缩放与发现性的相互作用。

Comments 58 Pages

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.06168 2026-01-13 cs.CV

Analyzing the Structure of Handwritten Digits: A Comparative Study of PCA, Factor Analysis, and UMAP

分析手写数字的结构:PCA、因子分析和UMAP的比较研究

Jyotiraditya Gupta

机构 * Department of Statistical Sciences(统计科学系) Department of Computer Science(计算机科学系) University of Toronto(多伦多大学)

AI总结 本文通过比较PCA、FA和UMAP,揭示手写数字在低维流形中的结构特性,展示不同方法对内在维度和几何结构的互补理解。

Comments 15 pages, 12 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.05600 2026-01-12 cs.CV cs.CL cs.LG

SceneAlign: Aligning Multimodal Reasoning to Scene Graphs in Complex Visual Scenes

SceneAlign: 在复杂视觉场景中将多模态推理对齐到场景图

Chuhan Wang, Xintong Li, Jennifer Yuntong Zhang, Junda Wu, Chengkai Huang, Lina Yao, Julian McAuley, Jingbo Shang

机构 * University of California, San Diego(加州大学圣地亚哥分校) University of Toronto(多伦多大学) University of New South Wales(新南威尔士大学)

AI总结 SceneAlign通过利用场景图进行结构干预,提升多模态推理在复杂视觉场景中的准确性和忠实性。

Comments Preprint

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.04616 2026-01-09 cs.LG cs.AI

DeepHalo: A Neural Choice Model with Controllable Context Effects

DeepHalo:具有可控上下文效应的神经选择模型

Shuhan Zhang, Zhi Wang, Rui Gao, Shuang Li

机构 * The Chinese University of Hong Kong, Shenzhen(香港中文大学(深圳)) University of Toronto(多伦多大学) The University of Texas at Austin(德克萨斯大学奥斯汀分校)

AI总结 DeepHalo是一种能够控制上下文效应阶数的神经选择模型,通过显式控制交互阶数和原则性解释上下文效应,提升决策建模的可解释性和预测性能。

详情

展开后加载摘要…

URL PDF HTML 收藏