语言大模型 / LLM - arXivDaily 专题

2511.22891 2025-12-01 cs.AI cs.CL cs.LG 82%

ORION: Teaching Language Models to Reason Efficiently in the Language of Thought

ORION：教语言模型以高效的方式在思维语言中推理

Kumar Tanmay, Kriti Aggarwal, Paul Pu Liang, Subhabrata Mukherjee

机构 * Harvard University（哈佛大学）； Hippocratic AI（希波克拉底AI）； Massachusetts Institute of Technology（麻省理工学院）

专题命中效率与部署：language model(title)；preference optimization(abstract)；分类 cs.CL、cs.AI、cs.LG

AI总结 ORION通过SLPO优化实现高效压缩推理，提升推理效率和准确性，同时保持高精度。

2511.21862 2025-12-01 cs.DC 82%

OOCO: Latency-disaggregated Architecture for Online-Offline Co-locate LLM Serving

OOCO：用于在线-离线协同LLM服务的延迟分解架构

Siyu Wu, Zihan Tang, Yuting Zeng, Hui Chen, Guiguang Ding, Tongxuan Liu, Ke Zhang, Hailong Yang

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)

AI总结 OOCO提出一种延迟分解架构，通过分离集群资源为严格和宽松延迟池，解决在线-离线协同LLM服务中的负载不平衡问题，提升离线吞吐量3倍并保持在线性能。

2511.18483 2025-11-25 cs.CY math.OC stat.AP 82%

Optimal Meal Schedule for a Local Nonprofit Using LLM-Aided Data Extraction

为本地非营利组织制定最优餐食计划：利用大语言模型辅助数据提取

Sergio Marin, Nhu Nguyen, Max, Zheng, Christina M. Weaver

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)

AI总结利用大语言模型辅助数据提取，为本地非营利组织制定最优餐食计划，通过二元整数规划和营养约束优化，实现成本最小化和营养均衡。

Comments 12 pages, 4 figures, presented at 2025 INFORMS Data Science Workshop (Atlanta, Georgia, Oct. 25, 2025)

2511.16467 2025-11-21 cs.CL cs.AI cs.LG 82%

Anatomy of an Idiom: Tracing Non-Compositionality in Language Models

成语的解剖：追踪语言模型中的非组合性

Andrew Gomes

机构 * Theoretical Particle Physics Laboratory, Institute of Physics, EPFL, Lausanne, Switzerland（理论粒子物理实验室，物理研究所，EPFL，洛桑，瑞士）

专题命中效率与部署：language model(title,abstract)；分类 cs.CL、cs.AI、cs.LG

AI总结本文研究了语言模型处理习语的非组合性机制，通过电路发现技术揭示了习语处理中的注意力模式和增强接收现象。

2501.02584 2025-11-21 cs.CV cs.AI cs.CL cs.LG 82%

Efficient Architectures for High Resolution Vision-Language Models

高效高分辨率视觉-语言模型架构

Miguel Carvalho, Bruno Martins

机构 * INESC-ID and Instituto Superior Técnico, University of Lisbon（INESC-ID 和里斯本大学技术学院）

专题命中效率与部署：language model(title,abstract)；分类 cs.CL、cs.AI、cs.LG

AI总结 Pheye是一种高效处理高分辨率图像的视觉-语言模型架构，通过减少参数数量实现高效率和强性能，尤其在细粒度图像理解和场景文本处理任务中表现突出。

Comments Accepted at COLING 2025

2511.12079 2025-11-20 cs.CV 82%

Point Cloud Quantization through Multimodal Prompting for 3D Understanding

Hongxuan Li, Wencheng Zhu, Huiying Xu, Xinzhong Zhu, Pengfei Zhu

专题命中效率与部署：prompting(title,abstract)；language model(abstract)

Comments Accepted by AAAI 2026. 11 pages, 7 figures

2511.11971 2025-11-18 physics.optics 82%

Channel Ordering for Fairness in Elastic Optical Networks via a LLM-Guided Bottleneck TSP Solver

Liangshun Wu, Wen Chen, Qingqing Wu

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)

2511.11356 2025-11-17 cs.CR 82%

SEAL: Subspace-Anchored Watermarks for LLM Ownership

Yanbo Dai, Zongjie Li, Zhenlan Ji, Shuai Wang

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)

2511.10692 2025-11-17 cs.SD 82%

StyleBreak: Revealing Alignment Vulnerabilities in Large Audio-Language Models via Style-Aware Audio Jailbreak

Hongyi Li, Chengxuan Zhou, Chu Wang, Sicheng Liang, Yanting Chen, Qinlin Xie, Jiawei Ye, Jie Wu

机构 * Hongyi Li, Chengxuan Zhou, Chu Wang, Sicheng Liang, Yanting Chen, Qinlin Xie, Jiawei Ye, Jie Wu（作者团队）

专题命中效率与部署：language model(title,abstract)；large language model(abstract)

Comments Accepted by AAAI 2026

2511.08903 2025-11-14 cs.CV 82%

LLM-Guided Probabilistic Fusion for Label-Efficient Document Layout Analysis

Ibne Farabi Shihab, Sanjeda Akter, Anuj Sharma

机构 * Department of Computer Science, Iowa State University（计算机科学系，爱荷华州立大学）； Department of Civil, Construction & Environmental Engineering, Iowa State University（土木、建设与环境工程系，爱荷华州立大学）

专题命中效率与部署：LLM(title,abstract)；pretraining(abstract)

2511.04477 2025-11-07 cs.DC 82%

Enabling Dynamic Sparsity in Quantized LLM Inference

Rongxiang Wang, Kangyuan Shu, Felix Xiaozhu Lin

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)

2511.03844 2025-11-07 cs.MA 82%

ASAP: an Agentic Solution to Auto-optimize Performance of Large-Scale LLM Training

Yuran Ding, Xinwei Chen, Xiaofan Zhang, Zongwei Zhou

专题命中效率与部署：LLM(title,abstract)；language model(abstract)

Comments This work has been accepted to Workshop on ML for Systems at NeurIPS 2025

2509.15433 2025-11-06 cs.CR 82%

LLM-Driven SAST-Genius: A Hybrid Static Analysis Framework for Comprehensive and Actionable Security

Vaibhav Agrawal, Kiarash Ahi

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)

Journal ref IEEE S&P 2025

2510.27037 2025-11-03 cs.CL cs.AI cs.LG cs.NE 82%

Elastic Architecture Search for Efficient Language Models

Shang Wang

机构 * Indepentdent Researcher（独立研究者）

专题命中效率与部署：language model(title,abstract)；分类 cs.CL、cs.AI、cs.LG

Comments ICME 2025

2510.26913 2025-11-03 cs.DC 82%

FlowMesh: A Service Fabric for Composable LLM Workflows

Junyi Shen, Noppanat Wadlom, Lingfeng Zhou, Dequan Wang, Xu Miao, Lei Fang, Yao Lu

专题命中效率与部署：LLM(title,abstract)；RLHF(abstract)

2504.15983 2025-11-03 cs.CL cs.AI cs.LG cs.NE 82%

W-PCA Based Gradient-Free Proxy for Efficient Search of Lightweight Language Models

Shang Wang

机构 * ShanghaiTech University（上海科技大学）

专题命中效率与部署：language model(title,abstract)；分类 cs.CL、cs.AI、cs.LG

Comments ICLR 2025

2510.21184 2025-10-27 cs.LG cs.AI cs.CL stat.ML 82%

Reducing the Probability of Undesirable Outputs in Language Models Using Probabilistic Inference

Stephen Zhao, Aidan Li, Rob Brekelmans, Roger Grosse

专题命中效率与部署：language model(title,abstract)；分类 cs.CL、cs.AI、cs.LG

2510.19307 2025-10-23 cs.CV 82%

Unified Reinforcement and Imitation Learning for Vision-Language Models

Byung-Kwan Lee, Ryo Hachiuma, Yong Man Ro, Yu-Chiang Frank Wang, Yueh-Hua Wu

机构 * NVIDIA ； KAIST（韩国科学技术院）； National Taiwan University（国立台湾大学）

专题命中效率与部署：language model(title,abstract)；LLM(abstract)

Comments NeurIPS 2025, Project page: https://byungkwanlee.github.io/RIL-page

2412.01822 2025-10-23 cs.CV 82%

VLsI: Verbalized Layers-to-Interactions from Large to Small Vision Language Models

Byung-Kwan Lee, Ryo Hachiuma, Yu-Chiang Frank Wang, Yong Man Ro, Yueh-Hua Wu

机构 * NVIDIA ； KAIST（韩国科学技术院）； National Taiwan University（国立台湾大学）

专题命中效率与部署：language model(title,abstract)；instruction tuning(abstract)

Comments CVPR 2025, Project page: https://byungkwanlee.github.io/VLsI-page/

2510.18525 2025-10-22 cs.AR 82%

From Quarter to All: Accelerating Speculative LLM Decoding via Floating-Point Exponent Remapping and Parameter Sharing

Yushu Zhao, Yubin Qin, Yang Wang, Xiaolong Yang, Huiming Han, Shaojun Wei, Yang Hu, Shouyi Yin

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)

2510.09607 2025-10-20 cs.CV 82%

VITA-VLA: Efficiently Teaching Vision-Language Models to Act via Action Expert Distillation

Shaoqi Dong, Chaoyou Fu, Haihan Gao, Yi-Fan Zhang, Chi Yan, Chu Wu, Xiaoyu Liu, Yunhang Shen, Jing Huo, Deqiang Jiang, Haoyu Cao, Yang Gao, Xing Sun, Ran He, Caifeng Shan

机构 * Nanjing University（南京大学）； Tencent Youtu Lab（腾讯优图实验室）； CASIA（中国科学院自动化研究所）

专题命中效率与部署：language model(title,abstract)；pretraining(abstract)

Comments Homepage: https://ltbai.github.io/VITA-VLA/

2406.07467 2025-10-10 cs.SE 82%

LLM meets ML: Data-efficient Anomaly Detection on Unstable Logs

Fatemeh Hadadi, Qinghua Xu, Domenico Bianculli, Lionel Briand

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)

Comments Accepted by ACM Transactions on Software Engineering and Methodology (TOSEM) 2025

2510.03727 2025-10-07 cs.AI cs.CL cs.CV cs.LG 82%

Bridging the Gap Between Multimodal Foundation Models and World Models

Xuehai He

机构 * Computer Science and Engineering University of California, Santa Cruz（计算机科学与工程大学加州大学圣克ruz分校）

专题命中效率与部署：foundation model(title,abstract)；分类 cs.CL、cs.AI、cs.LG

Comments PhD thesis

2508.15884 2025-09-30 cs.CL cs.AI cs.LG 82%

Jet-Nemotron: Efficient Language Model with Post Neural Architecture Search

Yuxian Gu, Qinghao Hu, Shang Yang, Haocheng Xi, Junyu Chen, Song Han, Han Cai

机构 * NVIDIA

专题命中效率与部署：language model(title,abstract)；分类 cs.CL、cs.AI、cs.LG

Comments NeurIPS 2025

2509.23625 2025-09-30 cs.CV cs.AI cs.CL cs.LG 82%

RIV: Recursive Introspection Mask Diffusion Vision Language Model

YuQian Li, Limeng Qiao, Lin Ma

机构 * Meituan（美团）

专题命中效率与部署：language model(title,abstract)；分类 cs.CL、cs.AI、cs.LG

2509.15777 2025-09-30 cs.SE 82%

Revisiting Vulnerability Patch Localization: An Empirical Study and LLM-Based Solution

Haoran Xu, Chen Zhi, Junxiao Han, Xinkui Zhao, Jianwei Yin, Shuiguang Deng

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)

2409.19375 2025-09-29 cs.LG cs.AI cs.CL cs.CV cs.HC 82%

DOTA: Distributional Test-Time Adaptation of Vision-Language Models

Zongbo Han, Jialong Yang, Guangyu Wang, Junfan Li, Qianli Xu, Mike Zheng Shou, Changqing Zhang

机构 * State Key Laboratory of Networking and Switching Technology（网络与交换技术国家重点实验室）； Beijing University of Posts and Telecommunications（北京邮电大学）； College of Intelligence and Computing（智能与计算学院）； Tianjin University（天津大学）； School of Computer Science and Technology（计算机科学与技术学院）； Harbin Institute of Technology Shenzhen（哈尔滨工业大学深圳学院）； Institute for Infocomm Research, A*STAR（信息通信研究机构，A*STAR）； Show Lab, National University of Singapore（新加坡国立大学Show实验室）； National University of Singapore（新加坡国立大学）

专题命中效率与部署：language model(title)；foundation model(abstract)；分类 cs.CL、cs.AI、cs.LG

Comments Accepted by NeurIPS 2025

2509.21301 2025-09-26 cs.OS 82%

Nova: Real-Time Agentic Vision-Language Model Serving with Adaptive Cross-Stage Parallelization

Yuhang Xu, Shengzhong Liu, Dong Zhang, Bingheng Yan, Fan Wu, Guihai Chen

专题命中效率与部署：language model(title,abstract)；LLM(abstract)

2509.16995 2025-09-23 cs.DC 82%

MoA-Off: Adaptive Heterogeneous Modality-Aware Offloading with Edge-Cloud Collaboration for Efficient Multimodal LLM Inference

Zheming Yang, Qi Guo, Yunqing Hu, Chang Zhao, Chang Zhang, Jian Zhao, Wen Ji

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)

Comments 5 pages, 4 figures

2508.05084 2025-09-15 cs.CV 82%

AdaFusion: Prompt-Guided Inference with Adaptive Fusion of Pathology Foundation Models

Yuxiang Xiao, Yang Hu, Bin Li, Tianyang Zhang, Zexi Li, Huazhu Fu, Jens Rittscher, Kaixiang Yang

专题命中效率与部署：foundation model(title,abstract)；pretraining(abstract)

Comments 6 Tables, 11 Figures

1. 效率与部署 22497 篇

ORION: Teaching Language Models to Reason Efficiently in the Language of Thought

OOCO: Latency-disaggregated Architecture for Online-Offline Co-locate LLM Serving

Optimal Meal Schedule for a Local Nonprofit Using LLM-Aided Data Extraction

Anatomy of an Idiom: Tracing Non-Compositionality in Language Models

Efficient Architectures for High Resolution Vision-Language Models

Point Cloud Quantization through Multimodal Prompting for 3D Understanding

Channel Ordering for Fairness in Elastic Optical Networks via a LLM-Guided Bottleneck TSP Solver

SEAL: Subspace-Anchored Watermarks for LLM Ownership

StyleBreak: Revealing Alignment Vulnerabilities in Large Audio-Language Models via Style-Aware Audio Jailbreak

LLM-Guided Probabilistic Fusion for Label-Efficient Document Layout Analysis

Enabling Dynamic Sparsity in Quantized LLM Inference

ASAP: an Agentic Solution to Auto-optimize Performance of Large-Scale LLM Training

LLM-Driven SAST-Genius: A Hybrid Static Analysis Framework for Comprehensive and Actionable Security

Elastic Architecture Search for Efficient Language Models

FlowMesh: A Service Fabric for Composable LLM Workflows

W-PCA Based Gradient-Free Proxy for Efficient Search of Lightweight Language Models

Reducing the Probability of Undesirable Outputs in Language Models Using Probabilistic Inference

Unified Reinforcement and Imitation Learning for Vision-Language Models

VLsI: Verbalized Layers-to-Interactions from Large to Small Vision Language Models

From Quarter to All: Accelerating Speculative LLM Decoding via Floating-Point Exponent Remapping and Parameter Sharing

VITA-VLA: Efficiently Teaching Vision-Language Models to Act via Action Expert Distillation

LLM meets ML: Data-efficient Anomaly Detection on Unstable Logs

Bridging the Gap Between Multimodal Foundation Models and World Models

Jet-Nemotron: Efficient Language Model with Post Neural Architecture Search

RIV: Recursive Introspection Mask Diffusion Vision Language Model

Revisiting Vulnerability Patch Localization: An Empirical Study and LLM-Based Solution

DOTA: Distributional Test-Time Adaptation of Vision-Language Models

Nova: Real-Time Agentic Vision-Language Model Serving with Adaptive Cross-Stage Parallelization

MoA-Off: Adaptive Heterogeneous Modality-Aware Offloading with Edge-Cloud Collaboration for Efficient Multimodal LLM Inference

AdaFusion: Prompt-Guided Inference with Adaptive Fusion of Pathology Foundation Models