arXivDaily arXiv每日学术速递 周一至周五更新

AI 大模型

语言大模型 / LLM

大语言模型、预训练、指令微调、后训练和语言模型应用。

2026-06-26 至 2026-06-26 共收录 23 信号源:cs.CL, cs.AI, cs.LG

1. 领域大模型 23 篇

2606.27047 2026-06-26 cs.CL cs.AI 新提交 93%

NuclearQAv2: A Structured Benchmark for Evaluating Domain-Science Competence in Large Language Models

NuclearQAv2:评估大型语言模型领域科学能力的结构化基准

Henry Shaowu Yuchi, Michal Kucer, Benjamin H. Sims, Selma Peterson, Emily Taylor

机构 * Los Alamos National Laboratory(洛斯阿拉莫斯国家实验室)

专题命中 领域大模型 :LLM(summary_cn,abstract);large language model(title,abstract);language model(title,abstract);prompting(abstract)

AI总结 提出NuclearQAv2基准,包含约1240个核工程问答对,通过混合管道生成,评估LLM在事实、数值和概念推理上的表现,发现定量推理仍具挑战。

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.26173 2026-06-26 cs.AI 新提交 92%

AlgoEvolve: LLM-driven Meta-evolution of Algorithmic Trading Programs

AlgoEvolve:LLM驱动的算法交易程序元进化

Dhruv Sharma, Gautam Shroff

机构 * Indraprastha Institute of Information Technology, Delhi(德里英迪拉普拉萨信息技术学院)

专题命中 领域大模型 :LLM(title,title_cn);large language model(abstract);language model(abstract);分类 cs.AI

AI总结 提出AlgoEvolve框架,利用LLM作为语义变异算子进化交易策略,通过元进化外层循环优化提示,在噪声、非平稳的金融环境中实现自适应策略逻辑,优于人工设计指令。

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.27052 2026-06-26 cs.CY nlin.AO 新提交 91%

Human--LLM Collaboration Is Transforming Complexity Metrics in Scientific Texts

人类与LLM协作正在改变科学文本中的复杂性度量

R. Alexander Bentley, Blai Vidiella, Damian J. Ruck, Senjuti Dutta, Kai Li, Sergi Valverde

专题命中 领域大模型 :LLM(title,title_cn);large language model(abstract);language model(abstract)

AI总结 研究利用2010-2025年arXiv摘要,发现2023年后LLM使用激增导致词汇更替率上升,且LLM风格指数与复杂性度量(词汇量、Heap定律和Zipf定律指数)的正相关关系变平缓,表明人机混合语言生态系统的涌现特性发生变化。

Comments 8 pages, 5 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.26884 2026-06-26 cs.HC 新提交 90%

MedSWFlow: An Open-Source LLM Workflow for Drafting Medical Social Work Case Plans

MedSWFlow: 一个用于起草医务社会工作案例计划的开源LLM工作流

Yulin Mao, Shiyu Li, Shuping Song, Yuling Zhang, Yajun Song

专题命中 领域大模型 :LLM(title,title_cn);prompting(abstract)

AI总结 提出开源、模型无关的LLM工作流MedSWFlow,将案例计划任务分解为六个阶段,通过结构化提示生成可审查的评估表和服务计划,旨在辅助从业者起草案例计划。

Comments 26pages, 8tables, 2figuers

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.26346 2026-06-26 cs.AI 新提交 90%

How Do Tool-Augmented LLM Agents Perform on Real-World Energy Analytics Tasks?

工具增强的LLM智能体在真实世界能源分析任务中的表现如何?

David Akinpelu, Akintonde Abbas, Rereloluwa Alimi, Ayodeji Lana

机构 * Independent Researcher(独立研究员) Tume AI

专题命中 领域大模型 :LLM(title,title_cn);分类 cs.AI

AI总结 针对能源领域缺乏智能体评估基准的问题,构建了包含243个专家策划问题的测试环境,评估工具增强的LLM智能体在数据检索、知识解释和定量建模三类任务上的表现。

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.06078 2026-06-26 cs.CY cs.AI cs.CL 90%

Simulating Students with Large Language Models: A Review of Architecture, Mechanisms, and Role Modelling in Education with Generative AI

用大型语言模型模拟学生:生成式AI在教育中的架构、机制与角色建模综述

Luis Marquez-Carpintero, Alberto Lopez-Sellers, Miguel Cazorla

机构 * Institute for Computer Research University of Alicante(计算机研究所阿利坎特大学)

专题命中 领域大模型 :large language model(title,abstract);language model(title,abstract);LLM(abstract);分类 cs.CL、cs.AI

AI总结 本文综述了利用大型语言模型模拟学生行为在教育中的应用,探讨了其在学习者建模、教学评估和教师培训中的潜力与挑战。

Journal ref Computer Science Review 62 (2026) 101008

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.27231 2026-06-26 quant-ph physics.ins-det 新提交 89%

A hardware-safety-gated system for LLM-written native ARTIQ control code on a trapped-ion platform

基于硬件安全门控的LLM编写原生ARTIQ控制代码系统在离子阱平台上的应用

Duanyang Wang, Lu Qi, Yuanheng Xie, Norbert M. Linke, Kenneth R. Brown

专题命中 领域大模型 :LLM(title,title_cn)

AI总结 提出一种硬件安全门控系统,通过令牌授权机制确保LLM代理在离子阱实验中安全自主地编写和执行控制代码,并在实验中验证了其有效性和可移植性。

Comments 15 pages, 5 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.14340 2026-06-26 cs.SD 版本更新 89%

Refining Pseudo-Audio Prompts with Speech-Text Alignment for Text-Only Domain Adaptation in LLM-Based ASR

通过语音-文本对齐细化伪音频提示以实现基于LLM的ASR的纯文本领域适应

Ryo Magoshi, Takashi Maekaku, Yusuke Shinohara

机构 * Kyoto University, Japan(京都大学,日本) LY Corporation, Japan(LY公司,日本)

专题命中 领域大模型 :LLM(title,title_cn)

AI总结 本文提出通过语音-文本对齐细化伪音频提示的方法,以提升基于LLM的ASR在纯文本领域适应中的性能,实验表明该方法在整体错误率和词汇覆盖率上优于现有方法。

Comments Accepted at Interspeech 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.26728 2026-06-26 cs.AI cs.LG cs.MA 新提交 82%

Scientific discovery as meta-optimization: a combinatorial optimization case study

科学发现作为元优化:一个组合优化案例研究

Yuan-Hang Zhang, Chesson Sipling, Massimiliano Di Ventra

机构 * University of California San Diego(加州大学圣迭戈分校)

专题命中 领域大模型 :LLM(abstract,abstract_cn);large language model(abstract);language model(abstract);分类 cs.AI、cs.LG

AI总结 提出将科学研究形式化为元优化,通过共识目标聚合方法,在3-SAT算法发现中实现从O(N^2.51)到O(N^1.33)的规模改进,最大实例加速约67倍。

Comments 35 pages, 6 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.19852 2026-06-26 cs.CL cs.LG 新提交 82%

Prompt, Plan, Extract: Zero-Shot Agentic LLMs Workflows for Lung Pathology Extraction from Clinical Narratives

提示、规划、提取:用于从临床叙述中提取肺部病理学的零样本智能体LLM工作流

Aman Pathak, Cheng Peng, Mengxian Lyu, Ziyi Chen, Reema Solan, Sankalp Talankar, Yasir Khan, Hiren Mehta, Aokun Chen, Yi Guo, Yonghui Wu

机构 * Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida(健康结果与生物医学信息学系,医学院,佛罗里达大学) Division of Pulmonary, Critical Care and Sleep Medicine, Department of Medicine, College of Medicine, University of Florida(呼吸科、重症医学科和睡眠医学科,医学系,医学院,佛罗里达大学) College of Nursing, Florida State University(护理学院,佛罗里达州立大学)

专题命中 领域大模型 :LLM(title_cn);large language model(abstract);language model(abstract);分类 cs.CL、cs.LG

AI总结 提出零样本智能体工作流,利用开源大语言模型从肺切除病理报告中提取13个CAP字段,在无训练下达到0.893 Micro-F1,接近监督方法。

Comments 7 pages, 2 figures, 3 tables. Affiliations: (1) Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, USA (2) Division of Pulmonary, Critical Care and Sleep Medicine, Department of Medicine, College of Medicine, University of Florida, Gainesville, FL, USA (3) College of Nursing, Florida State University, Tallahassee, FL, USA

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.04238 2026-06-26 cs.CV 版本更新 82%

6 Fingers, 1 Kidney: Natural Adversarial Medical Images Reveal Critical Weaknesses of Vision-Language Models

6根手指,1个肾脏:自然对抗性医学图像揭示视觉语言模型的关键弱点

Leon Mayer, Piotr Kalinowski, Caroline Ebersbach, Marcel Knopp, Tim Rädsch, Evangelia Christodoulou, Annika Reinke, Fiona R. Kolbinger, Lena Maier-Hein

机构 * German Cancer Research Center (DKFZ) Heidelberg, Division of Intelligent Medical Systems(德国癌症研究中心(DKFZ)海德堡,智能医学系统部门) Medical Faculty, Heidelberg University(海德堡大学医学院) Faculty of Mathematics and Computer Science, Heidelberg University(海德堡大学数学与计算机科学学院) HIDSS4Health - Helmholtz Information and Data Science School for Health, Karlsruhe/Heidelberg(HIDSS4Health - 哈勃-马克斯信息与数据科学健康学院,卡尔斯鲁厄/海德堡) Helmholtz Imaging, German Cancer Research Center (DKFZ)(哈勃-马克斯成像,德国癌症研究中心(DKFZ)) Engineering Faculty, Heidelberg University(海德堡大学工程学院) School of Computation, Information and Technology, TUM(技术大学(TUM)计算、信息与技术学院) Weldon School of Biomedical Engineering, Purdue University(普渡大学韦尔登生物医学工程学院) Department of Visceral, Thoracic and Vascular Surgery, University Hospital and Faculty of Medicine Carl Gustav Carus, TUD Dresden University of Technology(visceral、胸腔和血管外科部门,技术大学(TUD)德累斯顿大学医院和医学院) National Center for Tumor Diseases (NCT), NCT Heidelberg, a partnership between DKFZ and University Hospital Heidelberg(肿瘤疾病国家中心(NCT),海德堡NCT,DKFZ与海德堡大学医院之间的合作) Heidelberg University Hospital, Surgical Clinic, Surgical AI Research Group(海德堡大学医院,外科诊所,外科人工智能研究组) Mohamed Bin Zayed University of Artificial Intelligence (MBZUAI), Abu Dhabi, UAE(Mohamed Bin Zayed人工智能大学(MBZUAI),阿布扎赫,阿拉伯联合酋长国)

专题命中 领域大模型 :language model(title,abstract);prompting(abstract)

AI总结 提出AdversarialAnatomyBench基准,测试25个视觉语言模型在罕见解剖变异上的表现,发现准确率从71%降至28%,且模型缩放和干预无法解决。

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.26481 2026-06-26 cs.CL cs.DL cs.IR 新提交 81%

Extracting Problem and Method Sentence from Scientific Papers: A Context-enhanced Transformer Using Formulaic Expression Desensitization

从科学论文中提取问题和方法句子:一种使用公式化表达脱敏的上下文增强Transformer

Yingyi Zhang, Chengzhi Zhang

专题命中 领域大模型 :LLM(abstract,abstract_cn);large language model(abstract);language model(abstract);分类 cs.CL

AI总结 针对科学论文中问题和方法句子提取任务,提出公式化表达脱敏的数据增强方法和上下文增强Transformer,在小型数据集上提升宏F1分数3.71%和2.67%。

Journal ref Scientometrics, 2024

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.26573 2026-06-26 cs.CY cs.HC 新提交 80%

Pingquanqi (Equalizer): A Cross-Domain Sociotechnical Framework for Human-Agent Interaction Governance

Pingquanqi(均衡器):人机交互治理的跨领域社会技术框架

Yu Wang

专题命中 领域大模型 :LLM(summary_cn,abstract)

AI总结 提出Pingquanqi框架,通过用户状态区分、贝叶斯止损、可控摩擦和透明度度量等组件,优化LLM代理交互成本,保护用户时间资源,兼顾企业经济效益。

Comments 31 pages

详情

展开后加载摘要…

URL PDF HTML 收藏
2604.03724 2026-06-26 cs.IR 版本更新 80%

Rank, Don't Generate: Statement-level Ranking for Explainable Recommendation

排序,而非生成:面向可解释推荐的语句级排序

Ben Kabongo, Arthur Satouf, Vincent Guigue

专题命中 领域大模型 :LLM(abstract,abstract_cn);large language model(abstract);language model(abstract)

AI总结 提出将可解释推荐形式化为语句级排序问题,通过从评论中提取解释性语句并排序,避免生成式幻觉,支持细粒度事实分析和标准化评估。

Comments 12 pages, 7 tables, 5 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.27023 2026-06-26 cs.LG cs.CL cs.CV 新提交 79%

Just how sure are you? Improving Verbalized Uncertainty Calibration in Medical VQA

你到底有多确定?改进医学视觉问答中的口头不确定性校准

Eren Senoglu, Federico Toschi, Nicolo Brunello, Andrea Sassella, Mark James Carman

机构 * Politecnico di Milano(米兰理工大学)

专题命中 领域大模型 :large language model(abstract);language model(abstract);prompting(abstract);分类 cs.CL、cs.LG

AI总结 针对多模态大语言模型在医学VQA中过度自信的问题,提出基于复合损失函数的微调框架,通过校准项、锚定正则化、对比对齐和KL稳定项,将校准误差降低60%以上,判别力提升26%以上。

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.08420 2026-06-26 cs.CV 新提交 78%

CheXanatomy: Anatomy-Aware Vision-Language Modeling for Chest Radiographs

CheXanatomy: 面向胸部X光片的解剖感知视觉-语言建模

Sergios Gatidis, Curtis Langlotz, Christian Bluethgen

机构 * Stanford Center for Artificial Intelligence in Medicine and Imaging, Stanford University(斯坦福大学医学与影像人工智能中心) Department of Radiology, Stanford University(斯坦福大学放射学系)

专题命中 领域大模型 :language model(title,abstract)

AI总结 提出CheXanatomy框架,通过自回归令牌空间监督将解剖知识融入预训练视觉-语言模型,实现解剖分割,在合成和真实X光片上性能媲美U-Net,并提升域迁移鲁棒性和样本效率。

详情

展开后加载摘要…

URL PDF HTML 收藏
2508.03247 2026-06-26 cs.CL cs.CY 版本更新 77%

Somatic in the East, Psychological in the West?: Investigating Clinically-Grounded Cross-Cultural Depression Symptom Expression in LLMs

东方躯体化,西方心理化?:探究临床基础下跨文化抑郁症状在LLMs中的表达

Shintaro Sakai, Jisun An, Migyeong Kang, Haewoon Kwak

机构 * Indiana University Bloomington(印第安纳大学布卢明顿分校) Sungkyunkwan University(成均馆大学)

专题命中 领域大模型 :large language model(abstract);language model(abstract);prompting(abstract);分类 cs.CL

AI总结 本研究测试大语言模型(LLMs)是否复现西方抑郁患者报告心理症状、东方患者报告躯体症状的文化模式,发现英语提示下模型未能复现,但东方语言提示改善部分对齐,原因在于模型对文化人物敏感性低及症状层级压倒文化线索。

Comments C3NLP workshop at ACL 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.26118 2026-06-26 cs.CY cs.AI cs.LG 新提交 73%

The Open Source Economic Index of AI Adoption and Capability

AI采纳与能力的开源经济指数

Seamus Somerstep, Aritra Guha, Divesh Srivastava, Yuekai Sun

机构 * University of Michigan(密歇根大学) AT&T Chief Data Office(AT&T首席数据办公室) IFM-MBZUAI

专题命中 领域大模型 :LLM(abstract,abstract_cn);分类 cs.AI、cs.LG

AI总结 通过公开聊天数据和O*NET任务构建开源经济指数,衡量AI在各职业的采纳率与任务执行能力,发现金融、计算机和艺术行业采纳率最高。

详情

展开后加载摘要…

URL PDF HTML 收藏
2509.09960 2026-06-26 cs.LG cs.AI 版本更新 73%

Limited Reference, Reliable Generation: A Two-Component Framework for Tabular Data Generation in Low-Data Regimes

有限参考,可靠生成:低数据场景下的表格数据生成双组件框架

Mingxuan Jiang, Keyang Chen, Yongxin Wang, Yongsheng Zhao, Ziyue Dai, Yicun Liu, Zeping Li, Qiuyang Zhang, Hongyi Nie, Hongbin Zhu, Sen Liu, Guangnan Ye, Hongfeng Chai

机构 * School of Computer Science, Fudan University(复旦大学计算机科学学院) Institute of Financial Technology, Fudan University(复旦大学金融技术研究院) Northwestern Polytechnical University(西北工业大学)

专题命中 领域大模型 :large language model(abstract);language model(abstract);分类 cs.AI、cs.LG

AI总结 提出ReFine框架,通过从可解释模型提取符号规则嵌入提示引导生成,并采用双粒度过滤减少局部冗余,在低数据场景下实现稳健的表格数据生成,平均提升7.48%。

详情

展开后加载摘要…

URL PDF HTML 收藏
2506.18942 2026-06-26 cs.CY q-fin.RM 版本更新 67%

Advanced Applications of Generative AI in Actuarial Science: Case Studies Beyond ChatGPT

生成AI在精算科学中的高级应用:超越ChatGPT的案例研究

Simon Hatzesberger, Iris Nonneman

专题命中 领域大模型 :large language model(abstract);language model(abstract)

AI总结 本文通过四个案例研究探讨生成AI在精算实践中的潜力,涵盖从早期神经网络到现代Transformer生成AI系统的发展,展示其在索赔预测、市场比较、车辆损坏分类及代码迁移中的应用,同时讨论监管与安全挑战。

Comments v3: Minor revision in response to peer review. Methodological and reproducibility refinements across the case studies and the governance section

Journal ref European Actuarial Journal (2026)

详情

展开后加载摘要…

URL PDF HTML 收藏
2606.26448 2026-06-26 q-bio.NC cs.AI 新提交 57%

Closing the Loop to Discover Psychological Theories with an Automated Cognitive Scientist

闭环发现心理学理论:自动化认知科学家

Akshay K. Jagadish, Younes Strittmatter, Nori Jacoby, George Kachergis, Eric Schulz, Nathaniel Daw, Suyog H. Chandramouli, Thomas L. Griffiths

专题命中 领域大模型 :language model(abstract);分类 cs.AI

AI总结 提出AutoCog系统,利用大语言模型代理自动生成、测试和优化认知理论,在决策领域发现并验证了新理论。

Comments 44 pages, 9 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2604.26136 2026-06-26 eess.AS cs.CL 版本更新 57%

One Voice, Many Tongues: Cross-Lingual Voice Cloning for Scientific Speech

一个声音,多种语言:面向科学演讲的跨语言语音克隆

Amanuel Gizachew Abebe, Yasmin Moslem

机构 * Shaggar Institute of Technology(谢加尔技术学院) Trinity College Dublin(都柏林三一学院)

专题命中 领域大模型 :foundation model(abstract);分类 cs.CL

AI总结 本文针对跨语言科学演讲语音生成中的声音身份保持问题,提出基于OmniVoice基础模型的语音克隆系统,通过多模型集成蒸馏提升生成语音的可懂度和说话人相似性。

Comments In Proceedings of the 23rd International Conference on Spoken Language Translation (IWSLT 2026)

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.00813 2026-06-26 cs.CV 版本更新 50%

Generating a Paracosm for Training-Free Zero-Shot Composed Image Retrieval

生成一个Paracosm用于免训练零样本组合图像检索

Tong Wang, Yunhan Zhao, Shu Kong

机构 * University of Macau(澳门大学) UC Irvine(伊丽莎白女王大学) Institute of Collaborative Innovation(协同创新研究院)

专题命中 领域大模型 :language model(abstract)

AI总结 提出Paracosm方法,通过大多模态模型直接生成“心理图像”并构建合成域,实现免训练的零样本组合图像检索,在多个基准上达到最优性能。

Comments Accepted to ECCV 2026. Website and code: https://leowangtong.github.io/Paracosm/

详情

展开后加载摘要…

URL PDF HTML 收藏