arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

University of Texas at Austin(得克萨斯大学奥斯汀分校)

共收录 110
2512.25065 2026-06-17 cs.OS cs.AI cs.DC 版本更新

Vulcan: Instance-specialized, Verifiable Systems Heuristics Through LLM-driven Search

Vulcan:通过LLM驱动的搜索实现实例特化的可验证系统启发式方法

Rohit Dwivedula, Divyanshu Saxena, Sujay Yadalam, Eric Hayden Campbell, Daehyeok Kim, Aditya Akella

机构 * The University of Texas at Austin(德克萨斯大学奥斯汀分校)

AI总结 提出Vulcan框架,利用LLM生成系统启发式方法,通过隔离决策逻辑和受限语言Anvil保证安全,在调度、缓存和内存管理上取得显著性能提升。

Comments 19 pages

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.07082 2026-06-16 cs.LG cs.AI 版本更新

On the Geometry of On-Policy Distillation

论在线策略蒸馏的几何结构

Zhennan Shen, Yanshu Li, Qingyu Yin, Chak Tou Leong, Zhilin Wang, Yanxu Chen, Rongduo Han, Sunbowen Lee, Yi R. Fung

机构 * HKUST(香港科技大学) UT Austin(得克萨斯大学奥斯汀分校) Zhejiang University(浙江大学) Hong Kong PolyU(香港理工大学) USTC(中国科学技术大学) BUPT(北京邮电大学) Nankai University(南开大学) BIT(北京理工大学)

AI总结 本文通过参数空间诊断,揭示在线策略蒸馏(OPD)的更新轨迹具有松弛离主成分、子空间锁定等独特几何特性,表明其并非介于SFT和RLVR之间的中间方法。

Comments 17 pages, 8 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.12670 2026-06-16 cs.AI 版本更新

SkillsBench: Benchmarking How Well Agent Skills Work Across Diverse Tasks

SkillsBench: 基准测试智能体技能在不同任务中的有效性

Xiangyi Li, Yimin Liu, Wenbo Chen, Bingran You, Zonglin Di, Yifeng He, Shenghan Zheng, Kyoung Whan Choe, Jiankai Sun, Shuyi Wang, Chujun Tao, Binxu Li, Xuandong Zhao, Hejia Geng, Xiaojun Wu, Junwei Zhou, Xiaokun Chen, Hanwen Xing, Yubo Li, Qunhong Zeng, Di Wang, Yuanli Wang, Roey Ben Chaim, Penghao Jiang, Haotian Shen, Luyang Kong, Xinyi Liu, Runhui Wang, Xuanqing Liu, Jiachen Li, Xin Lan, Yueqian Lin, Wengao Ye, Junwei He, Songlin Li, Yue Zhang, Yipeng Gao, Yijiang Li, Ze Ma, Liqiang Jing, Tianyu Wang, Kaixin Li, Yiqi Xue, Haoran Lyu, Yizhuo He, Yuchen Tian, Shutong Wu, Bowei Wang, Yixuan Gao, Bo Chen, Litong Liu, Sikai Cheng, Jiajun Bao, Shuaicheng Tong, Shuwen Xu, Terry Yue Zhuo, Tinghan Ye, Qi Qi, Miao Li, Longtai Liao, Zelin Tan, Chang Shi, Xilin Tang, Srinath Tankasala, Boqin Yuan, Yaoyao Qian, Jianhong Tu, Chenguang Wang, Yizhou Sun, Wei Wang, Aaron Taylor, Ziyue Yang, Changkun Guan, Zhikang Dong, Xinyu Zhang, Steven Dillmann, Han-chung Lee, Dawn Song

机构 * BenchFlow OSU Amazon UC Berkeley UC Santa Cruz UC Davis Dartmouth RLWRLD Independent Princeton University Oxford University Stanford University USC CMU Foxconn Zenity UNSW UT Austin MSU Duke University ByteDance UT Dallas UC San Diego Columbia University University of Rochester Cornell Tech Georgia Tech Cornell University NEU UCLA Snap Inc. Fanshawe College University of Science and Technology of China HKUST(GZ) Anyscale

AI总结 提出SkillsBench基准,包含8领域87个任务,通过配对评估证明技能提升平均通过率16.6个百分点,小模型配备技能可匹敌大模型。

详情

展开后加载摘要…

URL PDF HTML 收藏
2410.00812 2026-06-16 cs.CL q-bio.NC 版本更新

Generative causal testing to bridge data-driven models and scientific theories in language neuroscience

生成式因果测试:弥合语言神经科学中数据驱动模型与科学理论之间的鸿沟

Richard Antonello, Chandan Singh, Shailee Jain, Aliyah Hsu, Sihang Guo, Jianfeng Gao, Bin Yu, Alexander Huth

机构 * Computer Science Department, University of Texas at Austin(德克萨斯大学计算机科学系) Microsoft Research(微软研究院) Neurosurgery Department, University of California(加州大学神经外科系) EECS Department, University of California(加州大学电子工程与计算机科学系) Statistics Department, University of California(加州大学统计学系) Center for Computational Biology, University of California(加州大学计算生物学中心) Neuroscience Department, University of California(加州大学神经科学系)

AI总结 提出生成式因果测试(GCT)框架,利用大语言模型生成简洁解释并通过LLM生成刺激进行验证,成功解释大脑区域的语言选择性,弥合数据驱动模型与科学理论之间的差距。

Comments Accepted to Nature Neuroscience, please cite that version

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.10886 2026-06-15 cs.CL 版本更新

MET-Bench: Multimodal Entity Tracking for Evaluating the Limitations of Vision-Language and Reasoning Models

MET-Bench:用于评估视觉语言与推理模型局限性的多模态实体追踪

Vanya Cohen, Raymond Mooney

机构 * University of Texas at Austin(德克萨斯大学奥斯汀分校)

AI总结 提出MET-Bench多模态实体追踪基准,发现视觉语言模型在图像实体追踪上显著弱于文本,主要源于视觉推理缺陷,强化学习可提升模态内性能但跨模态迁移不足。

Comments ICML 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.03120 2026-06-15 cs.LG cs.AI 版本更新

Quantized Evolution Strategies: High-precision Fine-tuning of Quantized LLMs at Low-precision Cost

量化进化策略:以低精度代价实现量化大语言模型的高精度微调

Yinggan Xu, Kajetan Schweighofer, Risto Miikkulainen, Xin Qiu

机构 * University of California, Los Angeles(加州大学洛杉矶分校) Cognizant AI Lab(Cognizant AI实验室) UT Austin(得克萨斯大学奥斯汀分校)

AI总结 提出量化进化策略(QES),通过集成累积误差反馈和无状态种子重放,直接在量化空间进行全参数微调,无需反向传播,显著优于现有零阶微调方法。

Comments Added more tasks and baselines

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.19314 2026-06-11 cs.AI cs.CL cs.LG 版本更新

PRInTS: Reward Modeling for Long-Horizon Information Seeking

PRInTS:面向长程信息检索的奖励建模

Jaewoo Lee, Archiki Prasad, Justin Chih-Yao Chen, Zaid Khan, Elias Stengel-Eskin, Mohit Bansal

机构 * University of North Carolina at Chapel Hill(北卡罗来纳大学教堂山分校) University of Texas at Austin(德克萨斯大学奥斯汀分校)

AI总结 提出PRInTS生成式过程奖励模型,通过密集评分和轨迹摘要提升长程信息检索中工具交互与推理能力,在多个基准上超越前沿模型。

Comments ACL 2026, 19 pages, code: https://github.com/G-JWLee/PRInTS

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.01248 2026-06-10 cs.LG 版本更新

$S^3$-R1: Learning to Retrieve and Answer Step-by-Step with Synthetic Data

$S^3$-R1: 通过合成数据学习逐步检索与回答

Harsh Goel, Akhil Udathu, Susmija Jabbireddy, Pradnesh Kalkar, Atharva Parulekar

机构 * The University of Texas at Austin(德克萨斯大学奥斯汀分校) Google DeepMind(谷歌DeepMind)

AI总结 提出S^3-R1框架,通过合成数据生成和密集奖励信号,解决强化学习后训练中稀疏奖励和缺乏多跳问题数据的问题,提升模型搜索与问答能力。

Comments Under Review

详情

展开后加载摘要…

URL PDF HTML 收藏
2604.20048 2026-06-10 cs.CL cs.CY 版本更新

Culturally uneven urban perception in large language models

大型语言模型通过文化不平等的基线感知城市

Rong Zhao, Wanqi Liu, Zhizhou Sha, Nanxi Su, Yecheng Zhang, Ying Long

机构 * Centre for Advanced Spatial Analysis (CASA), UCL, London, UK(高级空间分析中心(CASA),伦敦大学学院,英国) School of Architecture, Tsinghua University, Beijing, China(清华大学建筑学院,北京,中国) Department of Computer Science, UT Austin, Austin, TX, USA(得克萨斯大学奥斯汀分校计算机科学系,奥斯汀,德克萨斯,美国)

AI总结 本研究通过全球平衡的街景样本测试前沿LLM的城市感知,发现中性提示实际上偏向欧美文化,且文化提示能改变情感评价但无法恢复人类语义多样性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.00384 2026-06-09 cs.AI cs.CL cs.CV cs.LG stat.CO 版本更新

VESTA: Visual Exploration with Statistical Tool Agents

VESTA: 基于统计工具代理的视觉探索

William Rudman, Abhishek Divekar, Kanishk Jain, Sebastian Joseph, Stella S. R. Offner, Matthew Lease, Kyle Mahowald, Greg Durrett, Junyi Jessy Li

机构 * The University of Texas at Austin(德克萨斯大学奥斯汀分校) New York University(纽约大学)

AI总结 提出VESTA框架,通过动态增长的工具集指导数据变换、假设驱动可视化和统计检验,提升视觉语言模型在复杂统计建模任务上的性能。

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.26078 2026-06-09 cs.LG 版本更新

Global Convergence of Wasserstein Policy Gradient for Entropy-Regularized Reinforcement Learning

Wasserstein策略梯度在熵正则化强化学习中的全局收敛性

Zhaoyu Zhu, Rui Gao, Shuang Li

机构 * Shanghai Jiao Tong University(上海交通大学) The University of Texas at Austin(德克萨斯大学奥斯汀分校) The Chinese University of Hong Kong, Shenzhen(香港中文大学(深圳))

AI总结 本文通过利用熵正则化强化学习的Bellman结构,证明了Wasserstein策略梯度(WPG)方法的全局收敛性,并建立了分布Polyak-Łojasiewicz条件。

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.14285 2026-06-09 eess.IV cs.LG 版本更新

ForcingDAS: Unified and Robust Data Assimilation via Diffusion Forcing

通过扩散强迫实现统一且稳健的数据同化:ForcingDAS

Yixuan Jia, Siyi Chen, Yida Pan, Xiao Li, Lianghe Shi, Chanyong Jung, Haijie Yuan, Ismail Alkhouri, Yue Cynthia Wu, Saiprasad Ravishankar, Jeffrey A Fessler, Qing Qu

机构 * University of Michigan(密歇根大学) University of California, Berkeley(加州大学伯克利分校) Stanford University(斯坦福大学) Massachusetts Institute of Technology(麻省理工学院) University of Texas at Austin(德克萨斯大学奥斯汀分校)

AI总结 本文提出ForcingDAS,一种基于扩散强迫的统一数据同化框架,能够捕捉长时序依赖并减少误差积累,同时在推理时无需重新训练即可实现滤波到平滑的全谱应用。

详情

展开后加载摘要…

URL PDF HTML 收藏
2604.08479 2026-06-09 cs.CL 版本更新

AI generates well-liked but templatic empathic responses

AI生成受欢迎但模板化的共情回应

Emma S. Gueorguieva, Hongli Zhan, Jina Suh, Javier Hernandez, Tatiana Lau, Junyi Jessy Li, Desmond C. Ong

机构 * Department of Psychology, The University of Texas at Austin(心理学系,德克萨斯大学奥斯汀分校) Department of Linguistics, The University of Texas at Austin(语言学系,德克萨斯大学奥斯汀分校) Department of Computer Science and Engineering, The University of Washington(计算机科学与工程系,华盛顿大学) Microsoft Research(微软研究院) Toyota Research Institute(丰田研究院)

AI总结 研究发现LLM生成的共情回应高度模板化,采用10种共情语言策略,覆盖81-92%的回应内容,而人类写作则更多样。

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.18493 2026-06-09 eess.IV cs.AI cs.CV 版本更新

SAGE: Shape-Adapting Gated Experts for Adaptive Histopathology Image Segmentation

SAGE:适应性组织病理图像分割的形状自适应门控专家

Gia Huy Thai, Hoang-Nguyen Vu, Anh-Minh Phan, Quang-Thinh Ly, Thi-Ngoc-Truc Nguyen, Nhat Ho

机构 * University of Science, VNU-HCM(越南国家大学科学学院) Trivita AI University of Technology, VNU-HCM(越南国家大学技术学院) Michigan State University, USA(美国密歇根州立大学) The University of Texas at Austin(德克萨斯大学奥斯汀分校)

AI总结 SAGE通过动态专家路由框架提升异构视觉网络中细胞形态变化的适应性,实现高精度分割与稳健泛化。

Comments Accepted to CVPR 2026 (Findings Track). Project Page: https://oxyzgiahuy.github.io/sage/

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.10172 2026-06-09 astro-ph.IM cs.AI 版本更新

Cosmo3DFlow: Wavelet Flow Matching for Spatial-to-Spectral Compression in Reconstructing the Early Universe

Cosmo3DFlow:用于重建早期宇宙的空间到光谱压缩的小波流匹配

Md. Khairul Islam, Zeyu Xia, Ryan Goudjil, Jialu Wang, Arya Farahi, Judy Fox

机构 * Department of Computer Science University of Virginia(计算机科学系弗吉尼亚大学) Department of Statistics and Data Sciences The University of Texas at Austin(统计与数据科学系德克萨斯大学奥斯汀分校) School of Data Science(数据科学学院)

AI总结 提出Cosmo3DFlow框架,结合3D离散小波变换与流匹配,通过空间到光谱压缩解决高维宇宙结构重建中的维度和稀疏性瓶颈,实现比扩散模型快46倍的采样速度。

Journal ref KDD '26: Proc. 32nd ACM SIGKDD Conf. on Knowledge Discovery and Data Mining V.2, 11153-11164 (2026)

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.12744 2026-06-09 stat.ML cs.LG math.ST stat.CO stat.ME stat.TH 版本更新

Dendrograms of Mixing Measures for Softmax-Gated Gaussian Mixture of Experts: Consistency Without Model Sweeps

混合测度的树状图用于Softmax门控高斯混合专家:无需模型扫描的一致性

Do Tien Hai, Trung Nguyen Mai, TrungTin Nguyen, Nhat Ho, Binh T. Nguyen, Christopher Drovandi

机构 * Faculty of Mathematics and Computer Science, University of Science, Ho Chi Minh City, Vietnam(越南胡志明市科学大学数学与计算机科学学院) Vietnam National University Ho Chi Minh City, Vietnam(越南胡志明市国家大学) Faculty of Information Technology, University of Science, Ho Chi Minh City, Vietnam(越南胡志明市科学大学信息技术学院) ARC Centre of Excellence for the Mathematical Analysis of Cellular Systems(细胞系统数学分析 excellence 中心) School of Mathematical Sciences, Queensland University of Technology, Brisbane City, Australia(昆士兰科技大学数学科学学院) Department of Statistics and Data Science, University of Texas at Austin, Austin, USA(德克萨斯大学奥斯汀分校统计与数据科学系)

AI总结 针对softmax门控高斯混合专家模型,提出基于Voronoi损失函数的统一统计框架,解决参数非可识别性和模型选择问题,并引入混合测度树状图实现一致且无需多尺寸训练的专家数选择。

Comments Do Tien Hai, Trung Nguyen Mai, and TrungTin Nguyen are co-first authors. In Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, AISTATS 2026 Spotlight, Acceptance rate 2.5% over 2102 submissions

详情

展开后加载摘要…

URL PDF HTML 收藏
2509.15494 2026-06-09 cs.LG physics.data-an 版本更新

Multi-resolution Enhancement for Full Spectrum Neural Representations

全频谱神经表示的多分辨率增强

Yuan Ni, Zhantao Chen, Shizhou Xu, Cheng Peng, Rajan Plumley, Chun Hong Yoon, Jana B. Thayer, Joshua J. Turner

机构 * Linac Coherent Light Source, SLAC National Accelerator Laboratory(直线相干光源,SLAC国家加速器实验室) Stanford Institute for Materials and Energy Sciences, Stanford University(斯坦福大学材料与能源科学研究所) Walker Department of Mechanical Engineering, The University of Texas at Austin(德克萨斯大学奥斯汀分校机械工程系) Department of Mathematics, University of California Davis(加州大学戴维斯分校数学系) Department of Physics, Carnegie Mellon University(卡内基梅隆大学物理系)

AI总结 提出WIEN-INR框架,通过分层增强网络在不同分辨率尺度上建模,提升小网络对多尺度结构和高频细节的表示能力,实现紧凑高保真表示。

详情

展开后加载摘要…

URL PDF HTML 收藏
2508.20734 2026-06-09 cs.CV 版本更新

CardioMorphNet: Cardiac Motion Prediction Using a Shape-Guided Bayesian Recurrent Deep Network

CardioMorphNet: 使用形状引导的贝叶斯循环深度网络进行心脏运动预测

Reza Akbari Movahed, Abuzar Rezaee, Arezoo Zakeri, Colin Berry, Edmond S. L. Ho, Ali Gooya

机构 * University of California, San Diego(加州大学圣地亚哥分校) University of Texas at Austin(德克萨斯大学奥斯汀分校)

AI总结 提出CardioMorphNet,一种基于循环变分自编码器和贝叶斯公式的3D心脏形状引导可变形配准框架,通过递归配准分割图避免强度相似性损失,在心脏运动估计中优于现有方法,并具有更低的不确定性。

Comments Published in Medical Image Analysis. Updated to match the final published version

Journal ref Medical Image Analysis, vol. 113, p. 104149, 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.07833 2026-06-09 cs.DC cs.AI cs.MA cs.OS 版本更新

Harmonia: End-to-End RAG Serving Optimization

Harmonia: 端到端RAG服务优化

Saurabh Agarwal, Bodun Hu, Luis Pabon, Myungjin Lee, Jayanth Srinivasa, Aditya Akella

机构 * UT Austin(德克萨斯大学奥斯汀分校) Cisco Research(思科研究) Cisco Systems(思科系统)

AI总结 提出Harmonia框架,通过灵活管道接口、异构感知部署和闭环运行时控制器,优化RAG服务,吞吐量提升2.04倍以上,SLO违规减少78.4%。

详情

展开后加载摘要…

URL PDF HTML 收藏
2411.06469 2026-06-09 cs.CL 版本更新

ClinicalBench: Can LLMs Beat Traditional ML Models in Clinical Prediction?

ClinicalBench: 大型语言模型能在临床预测中击败传统机器学习模型吗?

Canyu Chen, Jian Yu, Shan Chen, Che Liu, Zhongwei Wan, Shuang Zhou, Yuan Luo, Rui Zhang, Danielle Bitterman, Fei Wang, Kai Shu

机构 * Department of Computer Science Northwestern University Evanston USA(计算机科学系西北大学艾文斯顿美国) Department of Computer Science University of Texas at Austin Austin USA(计算机科学系德克萨斯大学奥斯汀美国) Boston Children's Hospital, Harvard Medical School Boston USA(波士顿儿童医院哈佛医学院波士顿美国) Department of Computer Science Imperial College London London UK(计算机科学系伦敦帝国学院伦敦英国) Department of Computer Science Ohio State University Columbus USA(计算机科学系俄亥俄州立大学哥伦布美国) Massachusetts General Hospital, Harvard Medical School Boston USA(麻省总医院哈佛医学院波士顿美国) Department of Preventive Medicine, Feinberg School of Medicine Northwestern University Chicago USA(预防医学系费因伯格医学院西北大学芝加哥美国) Division of Computational Health Sciences, Department of Surgery University of Minnesota Minneapolis USA(计算健康科学部外科部明尼苏达大学明尼阿波利斯美国) Department of Population Health Sciences, Weill Cornell Medicine Cornell University New York USA(流行病学与公共卫生系韦尔·科恩医学中心康奈尔大学纽约美国) Department of Computer Science Emory University Atlanta USA(计算机科学系埃默里大学亚特兰大美国) Northwestern University(西北大学) University of Texas at Austin(德克萨斯大学奥斯汀) Boston Children's Hospital, Harvard Medical School(波士顿儿童医院哈佛医学院) Imperial College London(伦敦帝国学院) Ohio State University(俄亥俄州立大学) Massachusetts General Hospital, Harvard Medical School(麻省总医院哈佛医学院) University of Minnesota(明尼苏达大学) Cornell University(康奈尔大学) Emory University(埃默里大学)

AI总结 构建ClinicalBench基准,通过三个临床预测任务比较14个通用和8个医学LLM与11个传统ML模型,发现LLM在临床预测上仍无法超越传统ML模型。

Comments Accepted to Proceedings of KDD 2026. The first two authors contributed equally. 12 pages for main paper, 62 pages including appendix. Project website: https://clinicalbench.github.io

详情

展开后加载摘要…

URL PDF HTML 收藏