arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

University of Southern California(南加州大学)

共收录 1288
2310.05866 2026-02-02 quant-ph cs.AI cs.LG

Generative quantum machine learning via denoising diffusion probabilistic models

通过去噪扩散概率模型实现生成量子机器学习

Bingzhi Zhang, Peng Xu, Xiaohui Chen, Quntao Zhuang

机构 * Department of Physics and Astronomy, University of Southern California, Los Angeles, California 90089, USA(物理与天文学系,南加州大学) Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, California 90089, USA(明斯赫电气与计算机工程系,南加州大学) Department of Statistics, University of Illinois at Urbana-Champaign, Champaign, Illinois 61820, USA(统计系,伊利诺伊大学厄巴纳-香槟分校) Department of Mathematics, University of Southern California, Los Angeles, California 90089, USA(数学系,南加州大学)

AI总结 本文提出量子去噪扩散概率模型(QuDDPM),通过引入中间训练任务和多层电路结构,实现高效训练的量子数据生成学习,适用于相关量子噪声模型、量子多体相和拓扑结构学习。

Comments 5+10 pages, 16 figures. PRL accepted version. Code available at: https://github.com/francis-hsu/quantgenmdl

Journal ref Phys. Rev. Lett. 132, 100602 (2024)

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.21948 2026-01-30 cs.CV

Deep Models, Shallow Alignment: Uncovering the Granularity Mismatch in Neural Decoding

深度模型,浅层对齐:揭示神经解码中的粒度不匹配

Yang Du, Siyuan Dai, Yonghao Song, Paul M. Thompson, Haoteng Tang, Liang Zhan

机构 * Dept. of Electrical & Computer Engineering, University of Pittsburgh, USA(宾夕法尼亚大学电气与计算机工程系) Dept. of Biomedical Engineering, Tsinghua University, China(清华大学生物医学工程系) Dept. of Neurology, University of Southern California, USA(美国南加州大学神经病学系) Dept. of Computer Science, University of Texas Rio Grande Valley, USA(德克萨斯理工大学里奥格兰德谷分校计算机科学系)

AI总结 本文提出浅层对齐方法,通过对比学习策略解决神经解码中的粒度不匹配问题,显著提升解码性能。

Comments 29 pages, 13 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.21919 2026-01-30 cs.AI cs.CL

Self-Compression of Chain-of-Thought via Multi-Agent Reinforcement Learning

通过多智能体强化学习实现链式思维的自我压缩

Yiqun Chen, Jinyuan Feng, Wei Yang, Meizhi Zhong, Zhengliang Shi, Rui Li, Xiaochi Wei, Yan Gao, Yi Wu, Yao Hu, Zhiqiang Pu, Jiaxin Mao

机构 * Renmin University of China(中国人民大学) Xiaohongshu Inc.(小红书公司) Institute of Automation,Chinese Academy of Sciences(中国科学院自动化研究所) University of Southern California(南加州大学) Shandong University(山东大学)

AI总结 通过多智能体强化学习实现链式思维的自我压缩,有效减少响应长度并提升准确性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.21301 2026-01-30 cs.LG stat.ML

Achieving $\varepsilon^{-2}$ Dependence for Average-Reward Q-Learning with a New Contraction Principle

实现平均回报Q学习的ε⁻²依赖性:一种新的收缩原理

Zijun Chen, Zaiwei Chen, Nian Si, Shengbo Wang

机构 * Department of Computer Science and Engineering, HKUST(香港科技大学计算机科学与工程系) Edwardson School of Industrial Engineering, Purdue University(普渡大学工业工程学院) Department of Industrial Engineering and Decision Analytics, HKUST(香港科技大学工业工程与决策分析系) Daniel J. Epstein Department of Industrial and Systems Engineering, USC(美国南加州大学工业与系统工程系)

AI总结 本文提出了一种新的收缩原理,通过可达性假设实现了平均回报Q学习的最优ε⁻²样本复杂性保证。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.21076 2026-01-30 cs.AI

Multi-modal Imputation for Alzheimer's Disease Classification

多模态缺失数据填补用于阿尔茨海默病分类

Abhijith Shaji, Tamoghna Chattopadhyay, Sophia I. Thomopoulos, Greg Ver Steeg, Paul M. Thompson, Jose-Luis Ambite

机构 * Information Sciences Institute(信息科学研究所) University of Southern California(美国南加州大学) University of California(加州大学) Stevens Neuroimaging and Informatics Institute(史蒂文斯神经影像与信息学研究所)

AI总结 本文提出利用条件去噪扩散概率模型填补缺失的DWI扫描,以提高多模态深度学习模型在阿尔茨海默病三类分类中的准确性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2509.15689 2026-01-30 eess.IV cs.SD eess.AS

Interpretable Modeling of Articulatory Temporal Dynamics from real-time MRI for Phoneme Recognition

基于实时MRI的发声时序动力学可解释建模用于音素识别

Jay Park, Hong Nguyen, Sean Foley, Jihwan Lee, Yoonjeong Lee, Dani Byrd, Shrikanth Narayanan

机构 * Signal Analysis and Interpretation Lab, University of Southern California(南加州大学信号分析与解释实验室) Department of Linguistics, University of Southern California(南加州大学语言学系)

AI总结 本文提出基于实时MRI的发声时序动力学建模方法,通过多特征融合提升音素识别准确率,发现舌头和嘴唇运动对识别至关重要。

详情

展开后加载摘要…

URL PDF HTML 收藏
2508.12216 2026-01-29 cs.CV

Splat Feature Solver

基于稀疏线性逆问题的特征提升求解器

Butian Xiong, Rong Liu, Kenneth Xu, Meida Chen, Andrew Feng

机构 * University of Southern California, Institute for Creative Technologies(南加州大学,创意技术研究所) University of Michigan, Ann Arbor(密歇根大学安娜堡分校)

AI总结 Splat Feature Solver通过统一的稀疏线性逆问题框架,高效解决3D场景中特征提升问题,实现高质量提升特征生成,优于现有基线方法。

Comments ICLR 2026 Accepted

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.20126 2026-01-29 cs.CL cs.AI cs.LG

Rewarding Intellectual Humility Learning When Not To Answer In Large Language Models

在大型语言模型中不回答时如何奖励智力谦逊学习

Abha Jha, Akanksha Mahajan, Ashwath Vaithinathan Aravindan, Praveen Saravanan, Sai Sailaja Policharla, Sonal Chaturbhuj Gehlot

机构 * University of Southern California(南加州大学) Los Angeles, United States of America(美国洛杉矶)

AI总结 本文提出通过可验证奖励强化学习促进大型语言模型的智力谦逊,通过调整回避奖励机制减少错误响应,提升模型可靠性。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.19388 2026-01-29 cs.RO

Judgelight: Trajectory-Level Post-Optimization for Multi-Agent Path Finding via Closed-Subwalk Collapsing

Judgelight: 多智能体路径寻找中的轨迹级后优化 via 关闭子轨迹压缩

Yimin Tang, Sven Koenig, Erdem Bıyık

机构 * University of Southern California(南加州大学) University of California Irvine(加州大学尔湾分校)

AI总结 Judgelight通过压缩智能体轨迹中的闭合子轨迹,提高多智能体路径寻找的轨迹质量,减少20%的解决方案成本。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.14417 2026-01-29 cs.CL

Quantifying Speaker Embedding Phonological Rule Interactions in Accented Speech Synthesis

量化口音语法规则在口音语音合成中的相互作用

Thanathai Lertpetchpun, Yoonjeong Lee, Thanapat Trachu, Jihwan Lee, Tiantian Feng, Dani Byrd, Shrikanth Narayanan

机构 * Signal Analysis and Interpretation Lab, University of Southern California(南加州大学信号分析与解释实验室) Thomas Lord Department of Computer Science, University of Southern California(南加州大学托马斯·劳德计算机科学系) Department of Linguistics, University of Southern California(南加州大学语言学系)

AI总结 本研究通过量化说话者嵌入与语音规则的相互作用,提出 PSR 度量标准,揭示口音生成中规则与身份的纠缠关系。

Comments Accepted to ICASSP2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2402.03578 2026-01-29 cs.MA cs.AI

LLM Multi-Agent Systems: Challenges and Open Problems

大语言模型多智能体系统:挑战与开放问题

Shanshan Han, Qifan Zhang, Weizhao Jin, Zhaozhuo Xu

机构 * University of California, Irvine, CA, USA(加州大学尔湾分校) University of Southern California, Los Angeles, CA, USA(南加州大学) Stevens Institute of Technology, Hoboken, NJ, USA(史蒂文斯理工学院)

AI总结 本文研究了多智能体系统在任务分配、推理增强和内存管理方面的挑战,并探讨其在区块链中的应用潜力。

详情

展开后加载摘要…

URL PDF HTML 收藏
2310.00488 2026-01-29 cs.LG cs.AI

Membership Privacy Risks of Sharpness Aware Minimization

sharpness aware minimization 的成员隐私风险

Young In Kim, Andrea Agiollo, Pratiksha Agrawal, Johannes O. Royset, Rajiv Khanna

机构 * Department of Computer Science(计算机科学系) Purdue University(普渡大学) Delft University of Technology(代尔夫特理工大学) Department of Industrial and Systems Engineering(工业与系统工程系) University of Southern California(南加州大学)

AI总结 本文发现SAM在提升泛化能力的同时加剧了成员隐私泄露,通过分析记忆分数和影响分数揭示其对非典型子模式的捕捉能力更强,从而在成员推断攻击中表现更差。

Comments accepted to iclr 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.18305 2026-01-28 cs.CV

DiVE-k: Differential Visual Reasoning for Fine-grained Image Recognition

DiVE-k:基于微细图像识别的差分视觉推理

Raja Kumar, Arka Sadhu, Ram Nevatia

机构 * University of Southern California(南加州大学) Meta

AI总结 DiVE-k通过利用模型自身top-k预测作为训练信号,提升细粒度图像识别的差分推理能力,显著优于现有方法。

Comments ICLR 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.03906 2026-01-27 cs.CV

From Filters to VLMs: Benchmarking Defogging Methods through Object Detection and Segmentation Performance

从滤波器到视觉语言模型:通过目标检测和分割性能评估去雾方法

Ardalan Aryashad, Parsa Razmara, Amin Mahjoub, Seyedarmin Azizi, Mahdi Salmani, Arad Firouzkouhi

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

AI总结 本文通过目标检测和分割性能评估,探讨了去雾方法在真实与合成环境中的有效性,揭示了视觉语言模型在恶劣天气下的应用潜力。

Comments Accepted at WACV 2026 Proceedings (Oral), 5th Workshop on Image, Video, and Audio Quality Assessment in Computer Vision, with a focus on VLM and Diffusion Models

详情

展开后加载摘要…

URL PDF HTML 收藏
2509.21447 2026-01-27 eess.AS cs.AI cs.CL

ARTI-6: Towards Six-dimensional Articulatory Speech Encoding

ARTI-6:迈向六维发音语音编码

Jihwan Lee, Sean Foley, Thanathai Lertpetchpun, Kevin Huang, Yoonjeong Lee, Tiantian Feng, Louis Goldstein, Dani Byrd, Shrikanth Narayanan

机构 * Signal Analysis and Interpretation Lab, University of Southern California(南加州大学信号分析与解释实验室) Department of Linguistics, University of Southern California(南加州大学语言学系)

AI总结 ARTI-6通过六维发音特征集和反向合成模型,实现了高效且自然的语音生成与反向建模。

Comments Accepted for ICASSP 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.13321 2026-01-27 cs.HC cs.AI cs.CL

Adjust for Trust: Mitigating Trust-Induced Inappropriate Reliance on AI Assistance

调整信任:缓解因信任导致的对AI帮助的不当依赖

Tejas Srinivasan, Jesse Thomason

机构 * University of Southern California(美国南加州大学)

AI总结 通过信任自适应干预减少AI依赖,提升决策准确性与人机协作效率

详情

展开后加载摘要…

URL PDF HTML 收藏
2412.00435 2026-01-27 cs.AI cs.HC cs.RO

Towards Real-time Adaptation of Embodied Agent in Human-Robot Collaboration

面向人机协作中具身代理的实时适应

Shipeng Liu, Boshen Zhang, Zhehui Huang

机构 * University of Southern California Department of Electrical(南加州大学电气与计算机工程系) University of Southern California Department of Computer Science Los Angeles CA USA(南加州大学计算机科学系洛杉矶加州美国) University of Southern California(南加州大学)

AI总结 本文提出MonTA框架,通过轻量级监控和高效适配器提升人机协作中具身代理的实时适应能力,实现更高效的协作性能。

Comments 13 pages, 7 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.17588 2026-01-27 cs.AI cs.CL

Intelligence Requires Grounding But Not Embodiment

智能需要具身但不需要身体

Marcus Ma, Shrikanth Narayanan

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

AI总结 本文提出智能需要基础性而非具身,通过定义智能的四个属性并论证非具身智能体可实现这些属性,从而得出结论:基础性是智能的必要条件。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.17211 2026-01-27 cs.CV

Structural Complexity of Brain MRI reveals age-associated patterns

脑部MRI的结构复杂性揭示年龄相关模式

Anzhe Cheng, Italo Ivo Lima Dias Pinto, Paul Bogdan

机构 * University of Southern California, Los Angeles, CA, USA(美国南加州大学) Instituto de Matemática e Estatística, Universidade de São Paulo, São Paulo, Brazil(巴西圣保罗大学数学与统计研究所)

AI总结 本研究通过分析脑部MRI数据,揭示了结构复杂性随年龄变化的规律,并展示了其在预测生物年龄方面的应用价值。

Comments accepted by icassp2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2508.06515 2026-01-27 cs.CV

BigTokDetect: A Clinically-Informed Vision-Language Modeling Framework for Detecting Pro-Bigorexia Videos on TikTok

BigTokDetect: 一种临床指导的视觉-语言建模框架,用于检测TikTok上促进大肌肉畸形行为的视频

Minh Duc Chu, Kshitij Pawar, Zihao He, Roxanna Sharifi, Ross Sonnenblick, Magdalayna Curry, Laura D'Adamo, Lindsay Young, Stuart B Murray, Kristina Lerman

机构 * USC Information Sciences Institute(USC信息科学研究所) Keck School of Medicine, USC(USC凯克医学院) Department of Clinical Psychology, Drexel University(德雷塞尔大学临床心理学系) Department of Psychiatry and Biobehavioral Sciences, UCLA(UCLA精神病学与生物行为科学系)

AI总结 BigTokDetect通过临床指导的视觉-语言模型,检测TikTok上促进大肌肉畸形行为的视频,建立了可扩展的有害内容缓解框架。

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.14932 2026-01-27 cs.AI

FOL-Traces: Verified First-Order Logic Reasoning Traces at Scale

FOL-Traces: 在大规模上验证的首阶逻辑推理轨迹

Isabelle Lee, Sarah Liaw, Dani Yogatama

机构 * USC(美国大学) Harvard University(哈佛大学)

AI总结 FOL-Traces是一个大规模验证的首阶逻辑推理轨迹数据集,用于严格评估结构逻辑推理,通过挑战性任务揭示模型在语法意识和推理过程忠实度上的不足。

详情

展开后加载摘要…

URL PDF HTML 收藏
2411.01642 2026-01-27 cs.LG hep-ph

Quantum Rationale-Aware Graph Contrastive Learning for Jet Discrimination

量子理性感知图对比学习用于喷注鉴别

Md Abrar Jahin, Md. Akmol Masud, M. F. Mridha, Nilanjan Dey, Zeyar Aung

机构 * University of Southern California(南加州大学) Jahangirnagar University(贾哈吉尔纳加尔大学) American International University-Bangladesh(孟加拉国美国国际大学) Techno International New Town(技术国际新镇) Khalifa University(卡利法大学)

AI总结 本文提出量子理性感知图对比学习框架,通过量子理性生成器提升喷注鉴别性能,在参数受限环境下实现77.5%的AUC分数。

Journal ref Transactions on Machine Learning Research (TMLR), 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.16549 2026-01-26 cs.AI

LLM is Not All You Need: A Systematic Evaluation of ML vs. Foundation Models for text and image based Medical Classification

LLM并非万能:对ML与基础模型在文本和图像医学分类中的系统评估

Meet Raval, Tejul Pandit, Dhvani Upadhyay

机构 * University of Southern California Los Angeles, USA(美国南加州大学) Dhirubhani Ambani University Gandhinagar, India(印度达尔布哈尼·阿班尼大学)

AI总结 本文系统评估了ML与基础模型在医学分类中的表现,发现传统ML模型在多数任务中表现优异,而PEFT方法的效果依赖于适应策略。

Comments 9 pages, 5 figures, 3 tables, paper accepted in AAIML'26 conference

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.01078 2026-01-26 cs.AI

SimWorld: An Open-ended Realistic Simulator for Autonomous Agents in Physical and Social Worlds

SimWorld:一种用于物理和社会世界中自主代理的开放式真实模拟器

Jiawei Ren, Yan Zhuang, Xiaokang Ye, Lingjun Mao, Xuhong He, Jianzhi Shen, Mrinaal Dogra, Yiming Liang, Ruixuan Zhang, Tianai Yue, Yiqing Yang, Eric Liu, Ryan Wu, Kevin Benavente, Rajiv Mandya Nagaraju, Muhammad Faayez, Xiyan Zhang, Dhruv Vivek Sharma, Xianrui Zhong, Ziqiao Ma, Tianmin Shu, Zhiting Hu, Lianhui Qin

机构 * UCSD(加州大学圣地亚哥分校) UVA(弗吉尼亚大学) UIUC(伊利诺伊大学香槟分校) JHU(约翰·霍普金斯大学) Purdue(Purdue 大学) PolyU USC(美国南加州大学) UMich(密歇根大学)

AI总结 SimWorld是一个基于Unreal Engine 5构建的开放式真实模拟器,旨在开发和评估LLM/VLM代理在复杂物理和社会环境中的能力,通过多代理配送任务验证其推理模式与局限性。

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
2601.14283 2026-01-22 cs.LG cs.AI

Beyond Affinity: A Benchmark of 1D, 2D, and 3D Methods Reveals Critical Trade-offs in Structure-Based Drug Design

超越亲和力:一种1D、2D和3D方法的基准测试揭示了基于结构的药物设计中的关键权衡

Kangyu Zheng, Kai Zhang, Jiale Tan, Xuehan Chen, Yingzhou Lu, Zaixi Zhang, Lichao Sun, Marinka Zitnik, Tianfan Fu, Zhiding Liang

机构 * Department of Computer Science Rensselaer Polytechnic Institute(计算机科学系伦塞拉尔理工学院) Department of Computer Science and Engineering Lehigh University(计算机科学与工程系莱斯大学) Department of Computer Science University of Southern California(计算机科学系南加州大学) Stanford Medicine Department of Pathology Stanford University(斯坦福医学部病理学系斯坦福大学) Princeton University(普林斯顿大学) Harvard Medical School(哈佛医学院) State Key Laboratory for Novel Software Technology at Nanjing University(南京大学新型软件技术国家重点实验室) Department of Computer Science and Engineering The Chinese University of Hong Kong(计算机科学与工程系香港中文大学)

AI总结 本文通过对比1D、2D和3D方法,揭示了基于结构的药物设计中不同算法的性能差异及关键权衡。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.01090 2026-01-22 cs.MA cs.AI cs.CY

Harm in AI-Driven Societies: An Audit of Toxicity Adoption on Chirper.ai

AI驱动社会中的危害:对Chirper.ai上毒性采用的审计

Erica Coppolillo, Luca Luceri, Emilio Ferrara

机构 * University of Southern California, Los Angeles, California(美国南加州大学) University of Calabria, Rende, Italy(意大利卡拉布里亚大学)

AI总结 研究通过分析Chirper.ai上AI代理的毒性行为,揭示了暴露于有害内容如何影响代理行为,并提出通过监控毒性暴露来减轻有害行为的风险。

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.05665 2026-01-22 cs.CL cs.CV

Interleaved Latent Visual Reasoning with Selective Perceptual Modeling

交错的潜在视觉推理与选择性感知建模

Shuai Dong, Siyuan Wang, Xingyu Liu, Chenglin Li, Haowen Hou, Zhongyu Wei

机构 * China University of Geosciences, Wuhan(中国地质大学(武汉)) Shanghai Innovation Institute(上海创新研究院) University of Southern California(南加州大学) Fudan University(复旦大学) Zhejiang University(浙江大学) Shanghai Jiao Tong University(上海交通大学)

AI总结 ILVR通过交错潜在视觉表示与文本生成,实现动态状态演变与精确感知建模的统一,提升多模态推理性能。

Comments 18 pages, 11 figures. Code available at https://github.com/XD111ds/ILVR

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.13649 2026-01-21 cs.CL cs.AI

Fairness or Fluency? An Investigation into Language Bias of Pairwise LLM-as-a-Judge

公平性还是流畅性?对配对LLM-as-a-judge语言偏见的调查

Xiaolin Zhou, Zheng Luo, Yicheng Gao, Qixuan Chen, Xiyang Hu, Yue Zhao, Ruishan Liu

机构 * University of Southern California(南加州大学) Arizona State University(亚利桑那州立大学)

AI总结 本文研究了LLM-as-a-judge在配对任务中的语言偏见,发现不同语言在性能和偏好上有显著差异,且语言偏见不能仅由困惑度解释。

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.13284 2026-01-21 cs.LG

Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning

通过校准感知强化学习平衡决策LLM中的分类与校准性能

Duygu Nur Yaldiz, Evangelia Spiliopoulou, Zheng Qi, Siddharth Varia, Srikanth Doss, Nikolaos Pappas

机构 * University of Southern California(南加州大学) AWS AI Labs(AWS AI实验室)

AI总结 本文提出一种校准感知强化学习方法,通过调整决策标记概率,在保持准确性的同时减少过度自信,提升决策LLM的校准性能。

详情

展开后加载摘要…

URL PDF HTML 收藏