语言大模型 / LLM - arXivDaily 专题

2509.14252 2025-10-08 cs.CL cs.AI 92%

LLM-JEPA: Large Language Models Meet Joint Embedding Predictive Architectures

Hai Huang, Yann LeCun, Randall Balestriero

机构 * Atlassian（Atlassian公司）； NYU（纽约大学）； Brown University（布朗大学）

专题命中预训练与数据：LLM(title,abstract)；large language model(title,abstract)；language model(title,abstract)；pretraining(abstract)

2502.17355 2025-10-08 cs.CL 90%

On Relation-Specific Neurons in Large Language Models

Yihong Liu, Runsheng Chen, Lea Hirlimann, Ahmad Dawar Hakimi, Mingyang Wang, Amir Hossein Kargaran, Sascha Rothe, François Yvon, Hinrich Schütze

专题命中预训练与数据：large language model(title,abstract)；language model(title,abstract)；LLM(abstract)；pretraining(abstract)

Comments EMNLP 2025

2505.14824 2025-10-08 cs.CL 85%

Tracing Multilingual Factual Knowledge Acquisition in Pretraining

Yihong Liu, Mingyang Wang, Amir Hossein Kargaran, Felicia Körner, Ercong Nie, Barbara Plank, François Yvon, Hinrich Schütze

机构 * Center for Information and Language Processing, LMU Munich（信息与语言处理中心，慕尼黑大学）； Munich Center for Machine Learning (MCML)（慕尼黑机器学习中心（MCML））； Bosch Center for Artificial Intelligence（博世人工智能中心）； Sorbonne Université, CNRS, ISIR, France（索邦大学，法国国家科学研究中心，ISIR）

专题命中预训练与数据：pretraining(title,abstract)；large language model(abstract)；language model(abstract)；分类 cs.CL

Comments EMNLP Findings 2025

2510.05309 2025-10-08 cs.LG 84%

Gamma Mixture Modeling for Cosine Similarity in Small Language Models

Kevin Player

专题命中预训练与数据：language model(title)；small language model(title)；分类 cs.LG

Comments 16 pages, 8 figures

2510.06151 2025-10-08 cs.LG cs.AI cs.HC 79%

LLMs as Policy-Agnostic Teammates: A Case Study in Human Proxy Design for Heterogeneous Agent Teams

Aju Ani Justus, Chris Baber

机构 * School of Computer Science, University of Birmingham（伯明翰大学计算机科学学院）

专题命中预训练与数据：LLM(abstract)；large language model(abstract)；language model(abstract)；分类 cs.AI、cs.LG

Comments This is a preprint of a paper presented at the \textit{European Conference on Artificial Intelligence (ECAI 2025)}. It is made publicly available for the benefit of the research community and should be regarded as a preprint rather than a formally reviewed publication

2510.05133 2025-10-08 cs.CL 70%

Characterizing Model Behavior Under Synthetic Data Training: An Empirical Study Across Scales and Mixing Ratios

Y. Du, G. Wu, G. Tang, W. Wang, Q. Fan

机构 * Independent Research Collaboration（独立研究者）

专题命中预训练与数据：large language model(abstract)；language model(abstract)；分类 cs.CL

Comments 17 pages. Technical report

2506.19823 2025-10-08 cs.LG cs.AI 62%

Persona Features Control Emergent Misalignment

Miles Wang, Tom Dupré la Tour, Olivia Watkins, Alex Makelov, Ryan A. Chi, Samuel Miserendino, Jeffrey Wang, Achyuta Rajaram, Johannes Heidecke, Tejal Patwardhan, Dan Mossing

机构 * OpenAI

专题命中预训练与数据：language model(abstract)；分类 cs.AI、cs.LG

2505.11004 2025-10-08 cs.CL cs.AI 62%

Illusion or Algorithm? Investigating Memorization, Emergence, and Symbolic Processing in In-Context Learning

Jingcheng Niu, Subhabrata Dutta, Ahmed Elshabrawy, Harish Tayyar Madabushi, Iryna Gurevych

机构 * Technical University of Darmstadt（德累斯顿技术大学）； UKP Lab（UKP实验室）； Mohamed bin Zayed University of Artificial Intelligence（穆罕默德·本·扎耶德人工智能大学）； The University of Bath（巴斯大学）

专题命中预训练与数据：language model(abstract)；分类 cs.CL、cs.AI

Comments TMLR

2510.05558 2025-10-08 cs.CV cs.LG 57%

Midway Network: Learning Representations for Recognition and Motion from Latent Dynamics

Christopher Hoang, Mengye Ren

机构 * New York University（纽约大学）

专题命中预训练与数据：pretraining(abstract)；分类 cs.LG

Comments Project page: https://agenticlearning.ai/midway-network/

2510.05177 2025-10-08 eess.IV cs.LG 57%

Adapting HFMCA to Graph Data: Self-Supervised Learning for Generalizable fMRI Representations

Jakub Frac, Alexander Schmatz, Qiang Li, Guido Van Wingen, Shujian Yu

机构 * VU Amsterdam（阿姆斯特丹自由大学）； TReNDS Center（TReNDS中心）； Amsterdam UMC（阿姆斯特丹医学中心）

专题命中预训练与数据：pretraining(abstract)；分类 cs.LG

2509.18424 2025-10-08 cs.SD cs.AI eess.AS 57%

Scattering Transformer: A Training-Free Transformer Architecture for Heart Murmur Detection

Rami Zewail

机构 * Computer Science & Engineering Department（计算机科学与工程系）

专题命中预训练与数据：foundation model(abstract)；分类 cs.AI

Comments This paper has been accepted for presentation at the 14th International Conference on Model and Data Engineering (MEDI 2025). The final authenticated Version of Record will be published by Springer in the Lecture Notes in Computer Science (LNCS) series

2509.17794 2025-10-08 cs.CL 57%

Learning to vary: Teaching LMs to reproduce human linguistic variability in next-word prediction

Tobias Groot, Salo Lacunes, Evgenia Ilia

机构 * University of Amsterdam（阿姆斯特丹大学）

专题命中预训练与数据：language model(abstract)；分类 cs.CL

Comments EMNLP UncertaiNLP Workshop 2025

1. 预训练与数据 12 篇

LLM-JEPA: Large Language Models Meet Joint Embedding Predictive Architectures

On Relation-Specific Neurons in Large Language Models

Tracing Multilingual Factual Knowledge Acquisition in Pretraining

Gamma Mixture Modeling for Cosine Similarity in Small Language Models

LLMs as Policy-Agnostic Teammates: A Case Study in Human Proxy Design for Heterogeneous Agent Teams

Characterizing Model Behavior Under Synthetic Data Training: An Empirical Study Across Scales and Mixing Ratios

Persona Features Control Emergent Misalignment

Illusion or Algorithm? Investigating Memorization, Emergence, and Symbolic Processing in In-Context Learning

Midway Network: Learning Representations for Recognition and Motion from Latent Dynamics

Adapting HFMCA to Graph Data: Self-Supervised Learning for Generalizable fMRI Representations

Scattering Transformer: A Training-Free Transformer Architecture for Heart Murmur Detection

Learning to vary: Teaching LMs to reproduce human linguistic variability in next-word prediction