语言大模型 / LLM - arXivDaily 专题

2302.12441 2023-05-24 cs.LG cs.CL 84%

MUX-PLMs: Data Multiplexing for High-throughput Language Models

Vishvak Murahari, Ameet Deshpande, Carlos E. Jimenez, Izhak Shafran, Mingqiu Wang, Yuan Cao, Karthik Narasimhan

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

2304.03208 2023-04-07 cs.LG cs.CL 84%

Cerebras-GPT: Open Compute-Optimal Language Models Trained on the Cerebras Wafer-Scale Cluster

Nolan Dey, Gurpreet Gosal, Zhiming, Chen, Hemant Khachane, William Marshall, Ribhu Pathria, Marvin Tom, Joel Hestness

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

Comments 13 pages main text, 16 pages appendix, 13 figures

2303.13824 2023-03-27 cs.CL cs.AI 84%

$k$NN Prompting: Beyond-Context Learning with Calibration-Free Nearest Neighbor Inference

Benfeng Xu, Quan Wang, Zhendong Mao, Yajuan Lyu, Qiaoqiao She, Yongdong Zhang

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

Comments ICLR 2023. Code is available at https://github.com/BenfengXu/KNNPrompting

Journal ref ICLR 2023

2302.03773 2023-02-09 cs.CL cs.LG 84%

What Matters In The Structured Pruning of Generative Language Models?

Michael Santacroce, Zixin Wen, Yelong Shen, Yuanzhi Li

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

2302.01496 2023-02-06 cs.CL cs.LG cs.SD eess.AS 84%

Efficient Domain Adaptation for Speech Foundation Models

Bo Li, Dongseong Hwang, Zhouyuan Huo, Junwen Bai, Guru Prakash, Tara N. Sainath, Khe Chai Sim, Yu Zhang, Wei Han, Trevor Strohman, Francoise Beaufays

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

2210.13838 2022-10-27 cs.CL cs.LG 84%

Multilingual Relation Classification via Efficient and Effective Prompting

Yuxuan Chen, David Harbecke, Leonhard Hennig

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

Comments EMNLP 2022

2210.13578 2022-10-26 cs.CL cs.IR cs.LG 84%

Speeding Up Question Answering Task of Language Models via Inverted Index

Xiang Ji, Yesim Sungu-Eryilmaz, Elaheh Momeni, Reza Rawassizadeh

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

2207.07061 2022-10-26 cs.CL cs.LG 84%

Confident Adaptive Language Modeling

Tal Schuster, Adam Fisch, Jai Gupta, Mostafa Dehghani, Dara Bahri, Vinh Q. Tran, Yi Tay, Donald Metzler

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

Comments NeurIPS 2022 (selected as Oral)

2210.03858 2022-10-11 cs.LG cs.CL 84%

AlphaTuning: Quantization-Aware Parameter-Efficient Adaptation of Large-Scale Pre-Trained Language Models

Se Jung Kwon, Jeonghoon Kim, Jeongin Bae, Kang Min Yoo, Jin-Hwa Kim, Baeseong Park, Byeongwook Kim, Jung-Woo Ha, Nako Sung, Dongsoo Lee

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

Comments Findings of EMNLP 2022

2109.08449 2022-07-29 cs.CL cs.LG 84%

General Cross-Architecture Distillation of Pretrained Language Models into Matrix Embeddings

Lukas Galke, Isabelle Cuber, Christoph Meyer, Henrik Ferdinand Nölscher, Angelina Sonderecker, Ansgar Scherp

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

Comments 8 pages as accepted for WCCI/IJCNN 2022 + 8 pages supplementary material

2207.06814 2022-07-15 cs.CL cs.AI 84%

BERTIN: Efficient Pre-Training of a Spanish Language Model using Perplexity Sampling

Javier de la Rosa, Eduardo G. Ponferrada, Paulo Villegas, Pablo Gonzalez de Prado Salas, Manu Romero, Marıa Grandury

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

Comments Published at Procesamiento del Lenguaje Natural

Journal ref Procesamiento del Lenguaje Natural, 68 (2022): 13-23

2206.01861 2022-06-07 cs.CL cs.LG 84%

ZeroQuant: Efficient and Affordable Post-Training Quantization for Large-Scale Transformers

Zhewei Yao, Reza Yazdani Aminabadi, Minjia Zhang, Xiaoxia Wu, Conglong Li, Yuxiong He

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

Comments 11 pages, 4 figures

2204.06625 2022-04-19 cs.CL cs.LG 84%

CAMERO: Consistency Regularized Ensemble of Perturbed Language Models with Weight Sharing

Chen Liang, Pengcheng He, Yelong Shen, Weizhu Chen, Tuo Zhao

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

Comments Proceedings of ACL 2022

2110.13711 2022-04-19 cs.LG cs.CL 84%

Hierarchical Transformers Are More Efficient Language Models

Piotr Nawrot, Szymon Tworkowski, Michał Tyrolski, Łukasz Kaiser, Yuhuai Wu, Christian Szegedy, Henryk Michalewski

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

2204.06514 2022-04-14 cs.LG cs.CL 84%

Scalable Training of Language Models using JAX pjit and TPUv4

Joanna Yoo, Kuba Perlin, Siddhartha Rao Kamalakara, João G. M. Araújo

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

Comments 5 pages, 4 figures

2102.10407 2022-03-31 cs.CV cs.AI cs.CL cs.MM 84%

VisualGPT: Data-efficient Adaptation of Pretrained Language Models for Image Captioning

Jun Chen, Han Guo, Kai Yi, Boyang Li, Mohamed Elhoseiny

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

2102.12459 2021-09-16 cs.CL cs.LG 84%

When Attention Meets Fast Recurrence: Training Language Models with Reduced Compute

Tao Lei

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

Journal ref EMNLP 2021

2109.02040 2021-09-07 cs.CL cs.CV cs.LG 84%

Data Efficient Masked Language Modeling for Vision and Language

Yonatan Bitton, Gabriel Stanovsky, Michael Elhadad, Roy Schwartz

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

Comments Accepted to Findings of EMNLP 2021

2001.08896 2020-11-18 cs.LG cs.CL stat.ML 84%

Compressing Language Models using Doped Kronecker Products

Urmish Thakker, Paul N. Whatmough, Zhi-Gang Liu, Matthew Mattina, Jesse Beu

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

Comments Link to Workshop (https://research.fb.com/programs/on-device-intelligence-workshop/)

Journal ref Presented at On-device Intelligence Workshop at Third Conference on Machine Learning and Systems (MLSys) 2020

2608.12239 2026-08-13 cs.CV cs.AI cs.MM 新提交 84%

HAMP-LIC: Hessian-Aware Mixed-Precision Post-Training Quantization for Learned Image Compression

HAMP-LIC：面向学习型图像压缩的 Hessian 感知混合精度后训练量化

Yuefeng Zhang

机构 * Beijing Institute of Computer Technology and Application（北京计算机技术及应用研究所）； School of Elite Engineering, Northwestern Polytechnical University（西北工业大学精英工程学院）

专题命中效率与部署：post-training(title,abstract)；分类 cs.AI

AI总结本文针对学习型图像压缩模型低比特部署的效率与精度问题，提出带四阶段优化的 HAMP-LIC 框架，实现最高4.85倍模型压缩且仅0.59%率失真损失，性能优于现有方法并消除跨平台编解码误差。

Comments Learned image compression, post-training quantization, mixed-precision quantization, Hessian-based sensitivity analysis, model compression

2603.00856 2026-03-03 cs.SE cs.AI 84%

PARCER as an Operational Contract to Reduce Variance, Cost, and Risk in LLM Systems

PARCER作为运营合同以减少LLM系统中的方差、成本和风险

Elzo Brito dos Santos Filho

机构 * cps.sp.gov.br（巴西圣保罗州公共事务厅）

专题命中效率与部署：LLM(title,comments)；large language model(abstract)；language model(abstract)；分类 cs.AI

AI总结 PARCER通过引入运营合同框架，旨在减少LLM系统中的方差、成本和风险，提升治理和可控性。

Comments 11 pages. Presents the PARCER framework for LLM governance. Appendix contains the full YAML operational contract

2503.04346 2026-01-13 cs.CL 84%

Adding Alignment Control to Language Models

向语言模型添加对齐控制

Wenhong Zhu, Weinan Zhang, Rui Wang

专题命中效率与部署：language model(title,abstract)；post-training(abstract)；分类 cs.CL

AI总结本文提出CLM模型，通过添加身份层实现语言模型的对齐控制，通过插值系数实现对齐的插值和外推，提升模型的可用性。

Comments I have changed the title of the paper and resubmitted a new one on arXiv titled "Flexible realignment of language models" (arXiv:2506.12704)

2505.18541 2025-05-27 cs.AI 84%

RoleRAG: Enhancing LLM Role-Playing via Graph Guided Retrieval

Yongjie Wang, Jonathan Leung, Zhiqi Shen

机构 * Alibaba-NTU Global e-Sustainability CorpLab (ANGEL)（阿里巴巴-国立科技大学全球可持续性公司实验室（ANGEL））； College of Computing and Data Science（计算与数据科学学院）

专题命中效率与部署：LLM(title,comments)；large language model(abstract)；language model(abstract)；分类 cs.AI

Comments A Retrieval-enhanced LLM Role-playing

2305.15051 2024-06-04 cs.CL 84%

A Monte Carlo Language Model Pipeline for Zero-Shot Sociopolitical Event Extraction

Erica Cai, Brendan O'Connor

专题命中效率与部署：language model(title,abstract)；prompting(abstract)；分类 cs.CL；instruction tuning(comments)

Comments Accepted at NeurIPS 2023 Workshop on Instruction Tuning and Instruction Following; oral presentation at New England Natural Language Processing, 2023; 17 pages of text including references and appendix

2608.23259 2026-08-25 cs.SE 新提交 83%

TianoForge: An Automated Bug Triage Approach for the TianoCore UEFI Firmware Development Community

TianoForge：面向TianoCore UEFI固件开发社区的自动化Bug分类方法

Nazanin Siavash, Terrance E. Boult, Armin Moin

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

AI总结针对TianoCore UEFI固件开发社区，提出基于GPT-LLM（结合或不结合RAG）的自动化Bug分类方法TianoForge，可将平均Bug分类时间缩短99.95%，显著提升软件维护效率。

Comments ACM International Workshop on Firmware Testing and Analysis (FTA) 2026, Co-located with ISSTA 2026

2608.20169 2026-08-25 cs.CL cs.AI cs.LG 版本更新 83%

Task-CoEvolve: Efficient Harness Optimization via Adaptive Validation Task Selection

Task-CoEvolve：通过自适应验证任务选择实现高效的智能体框架优化

Atsuyuki Miyai, Kiyoharu Aizawa, Toshihiko Yamasaki

机构 * The University of Tokyo（东京大学）

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

AI总结 Task-CoEvolve通过自适应验证任务选择，使LLM智能体框架与验证任务协同演化，在保持全集搜索最终性能的同时，减少80%评估次数，实现高效的智能体框架优化。

Comments v2: Fix typo in v1, Github: https://github.com/Agent4Science-UTokyo/Task-CoEvolve

2608.20492 2026-08-24 cs.CV 新提交 83%

Annotations as Rollouts: Efficient and Scalable Reinforcement Learning for Video MLLMs

以标注为回滚：面向视频多模态大语言模型的高效可扩展强化学习

Yunheng Li, Guohong Mu, Hao Li, Shengsheng Qian, Dingwen Zhang, Qibin Hou, Ming-Ming Cheng

机构 * Nankai University（南开大学）； Northwestern Polytechnical University（西北工业大学）； Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）； NKIARI

专题命中效率与部署：SFT(abstract,abstract_cn)；large language model(abstract)；language model(abstract)；post-training(abstract)

AI总结本文提出OraRL算法，将标注作为神谕回滚解耦优势估计，实现高效可扩展的视频MLLM强化学习，在多项视频感知基准上超越现有模型，解码速度大幅提升。

Comments Project page: https://orarl.github.io/

2608.11121 2026-08-12 stat.OT stat.ME 新提交 83%

Generative AI use in Statistical Research: A Literature Review and Code Generation Case Study

生成式AI在统计研究中的应用：文献综述与代码生成案例研究

Natalie Morosin, Adel Ahmadi Nadi, Michael P. Wallace

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

AI总结本研究通过案例研究分析了ChatGPT-5和ScholarAI在统计研究的文献综述与代码生成中的应用，发现其需专业人员提示监督，可作为研究工具但无法替代专业方法论知识。

2511.07885 2026-08-10 cs.DC cs.AI cs.CL cs.LG 版本更新 83%

Intelligence per Watt: Measuring Intelligence Efficiency of Local AI

每瓦智能：衡量本地AI的智能效率

Jon Saad-Falcon, Avanika Narayan, Hakki Orhun Akengin, J. Wes Griffin, Herumb Shandilya, Adrian Gamarra Lafuente, Medhya Goel, Rebecca Joseph, Shlok Natarajan, Etash Kumar Guha, Shang Zhu, Ben Athiwaratkun, John Hennessy, Azalia Mirhoseini, Christopher Ré

机构 * Stanford University（斯坦福大学）； Together AI

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

AI总结本文研究了本地AI在能源效率和性能上的表现，提出了一种统一的衡量指标IPW，展示了本地推理在重新分配需求方面的能力，并揭示了本地加速器的优化潜力。

2603.17723 2026-07-20 q-fin.GN 版本更新 83%

LR-Robot: A Unified Supervised Intelligent Framework for Real-Time Systematic Literature Reviews with Large Language Models

LR-Robot：一种结合大语言模型的实时系统性文献综述统一监督智能框架

Wei Wei, Jin Zheng, Zining Wang

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

AI总结本文提出LR-Robot框架，结合AI与专家监督，实现系统性文献综述的多维分类、关系映射和细粒度主题演化分析，通过期权定价案例展示其在文献综合中的应用与效果。

Comments I've uploaded an updated paper and now it's already on arXiv (arXiv:2604.14793)

1. 效率与部署 22405 篇

MUX-PLMs: Data Multiplexing for High-throughput Language Models

Cerebras-GPT: Open Compute-Optimal Language Models Trained on the Cerebras Wafer-Scale Cluster

$k$NN Prompting: Beyond-Context Learning with Calibration-Free Nearest Neighbor Inference

What Matters In The Structured Pruning of Generative Language Models?

Efficient Domain Adaptation for Speech Foundation Models

Multilingual Relation Classification via Efficient and Effective Prompting

Speeding Up Question Answering Task of Language Models via Inverted Index

Confident Adaptive Language Modeling

AlphaTuning: Quantization-Aware Parameter-Efficient Adaptation of Large-Scale Pre-Trained Language Models

General Cross-Architecture Distillation of Pretrained Language Models into Matrix Embeddings

BERTIN: Efficient Pre-Training of a Spanish Language Model using Perplexity Sampling

ZeroQuant: Efficient and Affordable Post-Training Quantization for Large-Scale Transformers

CAMERO: Consistency Regularized Ensemble of Perturbed Language Models with Weight Sharing

Hierarchical Transformers Are More Efficient Language Models

Scalable Training of Language Models using JAX pjit and TPUv4

VisualGPT: Data-efficient Adaptation of Pretrained Language Models for Image Captioning

When Attention Meets Fast Recurrence: Training Language Models with Reduced Compute

Data Efficient Masked Language Modeling for Vision and Language

Compressing Language Models using Doped Kronecker Products

HAMP-LIC: Hessian-Aware Mixed-Precision Post-Training Quantization for Learned Image Compression

PARCER as an Operational Contract to Reduce Variance, Cost, and Risk in LLM Systems

Adding Alignment Control to Language Models

RoleRAG: Enhancing LLM Role-Playing via Graph Guided Retrieval

A Monte Carlo Language Model Pipeline for Zero-Shot Sociopolitical Event Extraction

TianoForge: An Automated Bug Triage Approach for the TianoCore UEFI Firmware Development Community

Task-CoEvolve: Efficient Harness Optimization via Adaptive Validation Task Selection

Annotations as Rollouts: Efficient and Scalable Reinforcement Learning for Video MLLMs

Generative AI use in Statistical Research: A Literature Review and Code Generation Case Study

Intelligence per Watt: Measuring Intelligence Efficiency of Local AI

LR-Robot: A Unified Supervised Intelligent Framework for Real-Time Systematic Literature Reviews with Large Language Models