大模型推理能力 - arXivDaily 专题

2509.17784 2025-10-31 cs.LG cs.AI 62%

Revealing Multimodal Causality with Large Language Models

Jin Li, Shoujin Wang, Qi Zhang, Feng Liu, Tongliang Liu, Longbing Cao, Shui Yu, Fang Chen

机构 * University of Technology Sydney（技术大学悉尼大学）； Tongji University（同济大学）； University of Melbourne（墨尔本大学）； University of Sydney（悉尼大学）； Macquarie University（麦考瑞大学）

专题命中其他推理：reasoning(abstract)；分类 cs.AI、cs.LG

Comments Accepted at NeurIPS 2025

2510.25460 2025-10-30 cs.CL cs.AI 62%

Fine-Tuned Language Models for Domain-Specific Summarization and Tagging

Jun Wang, Fuming Lin, Yuyu Chen

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

2510.24760 2025-10-30 cs.CL cs.AI 62%

Dingtalk DeepResearch: A Unified Multi Agent Framework for Adaptive Intelligence in Enterprise Environments

Mengyuan Chen, Chengjun Dai, Xinyang Dong, Chengzhe Feng, Kewei Fu, Jianshe Li, Zhihan Peng, Yongqi Tong, Junshao Zhang, Hong Zhu

机构 * Industrial Brain Team, Dingtalk, Alibaba Group（钉钉工业大脑团队，钉钉，阿里巴巴集团）

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

2411.12118 2025-10-29 cs.LG cs.CL 62%

Mechanism and Emergence of Stacked Attention Heads in Multi-Layer Transformers

Tiberiu Musat

机构 * ETH Zürich（苏黎世联邦理工学院）

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.LG

Journal ref International Conference on Learning Representations (ICLR), 2025

2510.24285 2025-10-29 cs.CV cs.AI cs.CL 62%

ViPER: Empowering the Self-Evolution of Visual Perception Abilities in Vision-Language Model

Juntian Zhang, Song Jin, Chuanqi Cheng, Yuhan Liu, Yankai Lin, Xun Zhang, Yufei Zhang, Fei Jiang, Guojun Yin, Wei Lin, Rui Yan

机构 * Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学人工智能学院）； Meituan（美团）； MBZUAI ； Wuhan University（武汉大学）

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

2510.15870 2025-10-29 cs.CV cs.AI cs.CL 62%

OmniVinci: Enhancing Architecture and Data for Omni-Modal Understanding LLM

Hanrong Ye, Chao-Han Huck Yang, Arushi Goel, Wei Huang, Ligeng Zhu, Yuanhang Su, Sean Lin, An-Chieh Cheng, Zhen Wan, Jinchuan Tian, Yuming Lou, Dong Yang, Zhijian Liu, Yukang Chen, Ambrish Dantrey, Ehsan Jahangiri, Sreyan Ghosh, Daguang Xu, Ehsan Hosseini-Asl, Danial Mohseni Taheri, Vidya Murali, Sifei Liu, Yao Lu, Oluwatobi Olabiyi, Yu-Chiang Frank Wang, Rafael Valle, Bryan Catanzaro, Andrew Tao, Song Han, Jan Kautz, Hongxu Yin, Pavlo Molchanov

机构 * NVIDIA

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

Comments Technical Report. Code: https://github.com/NVlabs/OmniVinci

2510.23408 2025-10-28 cs.AI cs.DC cs.ET cs.LG cs.MA 62%

AutoStreamPipe: LLM Assisted Automatic Generation of Data Stream Processing Pipelines

Abolfazl Younesi, Zahra Najafabadi Samani, Thomas Fahringer

机构 * Departement of Computer Science, University of Innsbruck（因斯布鲁克大学计算机科学系）

专题命中其他推理：reasoning(abstract)；分类 cs.AI、cs.LG

Comments Under review

2510.22993 2025-10-28 cs.LG cs.CL 62%

Can Language Models Compose Skills In-Context?

Zidong Liu, Zhuoyan Xu, Zhenmei Shi, Yingyu Liang

机构 * The University of Hong Kong（香港大学）； University of Wisconsin-Madison（威斯康星大学麦迪逊分校）

专题命中其他推理：chain-of-thought(abstract)；分类 cs.CL、cs.LG

2510.22729 2025-10-28 cs.AI cs.CL 62%

Critical Insights into Leading Conversational AI Models

Urja Kohli, Aditi Singh, Arun Sharma

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

Comments 21 pages, 7 tables, 3 figures. Open-access preprint intended for journal or conference submission

2510.22679 2025-10-28 cs.AI cs.CL 62%

Do Stop Me Now: Detecting Boilerplate Responses with a Single Iteration

Yuval Kainan, Shaked Zychlinski

机构 * JFrog

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

Comments 13 pages, 4 figures

2510.22467 2025-10-28 cs.LG cs.AI 62%

Backward-Friendly Optimization: Training Large Language Models with Approximate Gradients under Memory Constraints

Jing Yang, Kaitong Cai, Yijia Fan, Yufeng Yang, Keze Wang

机构 * Sun Yat-sen University（中山大学）

专题命中其他推理：reasoning(abstract)；分类 cs.AI、cs.LG

2509.23659 2025-10-28 cs.CL cs.AI 62%

Aligning LLMs for Multilingual Consistency in Enterprise Applications

Amit Agarwal, Hansa Meghwani, Hitesh Laxmichand Patel, Tao Sheng, Sujith Ravi, Dan Roth

机构 * Oracle AI

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

Comments Accepted at EMNLP 2025

2508.02366 2025-10-28 cs.LG cs.CL q-fin.TR 62%

Language Model Guided Reinforcement Learning in Quantitative Trading

Adam Darmanin, Vince Vella

机构 * University of Malta（马耳他大学）

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.LG

Comments 12 pages (4 pages appendix and references) and 6 figures. Accepted for presentation at FLLM 2025, Vienna

2504.07830 2025-10-28 cs.CL cs.AI cs.SI 62%

MOSAIC: Modeling Social AI for Content Dissemination and Regulation in Multi-Agent Simulations

Genglin Liu, Vivian Le, Salman Rahman, Elisa Kreiss, Marzyeh Ghassemi, Saadia Gabriel

机构 * University of California, Los Angeles（加州大学洛杉矶分校）； MIT CSAIL（麻省理工学院计算机科学与人工智能实验室）

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

Comments Accepted into EMNLP 2025 Main Conference, Oral Presentation

2503.10635 2025-10-28 cs.CV cs.AI cs.LG 62%

A Frustratingly Simple Yet Highly Effective Attack Baseline: Over 90% Success Rate Against the Strong Black-box Models of GPT-4.5/4o/o1

Zhaoyi Li, Xiaohan Zhao, Dong-Dong Wu, Jiacheng Cui, Zhiqiang Shen

机构 * VILA Lab, Department of Machine Learning, MBZUAI（VILA实验室，机器学习系，MBZUAI）

专题命中其他推理：reasoning(abstract)；分类 cs.AI、cs.LG

Comments NeurIPS 2025. Code at: https://github.com/VILA-Lab/M-Attack

2509.24841 2025-10-27 cs.CL cs.AI 62%

A Hierarchical Error Framework for Reliable Automated Coding in Communication Research: Applications to Health and Political Communication

Zhilong Zhao, Yindi Liu

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

Comments Version 2: Enhanced clarification of precision-matching task characteristics and framework applicability conditions. 20 pages, 4 figures, 4 tables. Replication package available at https://doi.org/10.7910/DVN/NDXVLZ

2506.21584 2025-10-27 cs.CL cs.AI cs.CY 62%

Empirical Evidence for Alignment Faking in a Small LLM and Prompt-Based Mitigation Techniques

Jeanice Koorndijk

机构 * Seraphion Technology（塞拉菲昂技术）

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

Comments NeurIPS RegML Workshop

2505.15293 2025-10-24 cs.LG cs.AI 62%

LLM-Explorer: A Plug-in Reinforcement Learning Policy Exploration Enhancement Driven by Large Language Models

Qianyue Hao, Yiwen Song, Qingmin Liao, Jian Yuan, Yong Li

机构 * Department of Electronic Engineering, BNRist, Tsinghua University（电子工程系、北京理工大学、清华大学）

专题命中其他推理：reasoning(abstract)；分类 cs.AI、cs.LG

2510.19008 2025-10-23 cs.HC cs.AI cs.LG cs.MA 62%

Plural Voices, Single Agent: Towards Inclusive AI in Multi-User Domestic Spaces

Joydeep Chandra, Satyam Kumar Navneet

机构 * BNRIST, Tsinghua University（北京理工大学、清华大学）； Independent Researcher（独立研究者）

专题命中其他推理：reasoning(abstract)；分类 cs.AI、cs.LG

2510.18297 2025-10-22 cs.CL cs.AI 62%

From Retrieval to Generation: Unifying External and Parametric Knowledge for Medical Question Answering

Lei Li, Xiao Zhou, Yingying Zhang, Xian Wu

机构 * Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学北京校区人工智能学院）； Tencent Jarvis Lab（腾讯 Jarvis 实验室）

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

Comments 13 pages, 4 figures

2509.17999 2025-10-22 cs.CY cs.AI cs.HC cs.LG 62%

The Narcissus Hypothesis: Descending to the Rung of Illusion

Riccardo Cadei, Christian Internò

专题命中其他推理：reasoning(abstract)；分类 cs.AI、cs.LG

Comments NeurIPS 2025 Workshop on Evaluating the Evolving LLM Lifecycle: Benchmarks, Emergent Abilities, and Scaling

2510.17937 2025-10-22 cs.LG cs.AI 62%

UniRL-Zero: Reinforcement Learning on Unified Models with Joint Language Model and Diffusion Model Experts

Fu-Yun Wang, Han Zhang, Michael Gharbi, Hongsheng Li, Taesung Park

机构 * Cuhk（香港中文大学）

专题命中其他推理：reasoning(abstract)；分类 cs.AI、cs.LG

2510.17880 2025-10-22 cs.CL cs.AI 62%

Outraged AI: Large language models prioritise emotion over cost in fairness enforcement

Hao Liu, Yiqing Dai, Haotian Tan, Yu Lei, Yujia Zhou, Zhen Wu

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

2505.16984 2025-10-21 cs.LG cs.CL 62%

UFT: Unifying Supervised and Reinforcement Fine-Tuning

Mingyang Liu, Gabriele Farina, Asuman Ozdaglar

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.LG

2510.16276 2025-10-21 cs.AI cs.LG 62%

What Limits Agentic Systems Efficiency?

Song Bian, Minghao Yan, Anand Jayarajan, Gennady Pekhimenko, Shivaram Venkataraman

机构 * UW-Madison（威斯康星大学麦迪逊分校）； University of Toronto（多伦多大学）； NVIDIA（英伟达）

专题命中其他推理：reasoning(abstract)；分类 cs.AI、cs.LG

Comments 27 pages, 15 figures

2510.14961 2025-10-17 cs.LG cs.CL 62%

Efficient Parallel Samplers for Recurrent-Depth Models and Their Connection to Diffusion Language Models

Jonas Geiping, Xinyu Yang, Guinan Su

机构 * ELLIS Institute Tübingen & Max-Planck Institute for Intelligent Systems, Tübingen AI Center（图宾根ELLIS研究所及图宾根马克斯·普朗克智能系统研究所AI中心）

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.LG

Comments Code can be found at https://github.com/seal-rg/recurrent-pretraining

2510.13831 2025-10-17 cs.CL cs.AI 62%

Informed Routing in LLMs: Smarter Token-Level Computation for Faster Inference

Chao Han, Yijuan Liang, Zihao Xuan, Daokuan Wu, Wei Zhang, Xiaoyu Shen

机构 * Institute of Digital Twin, Eastern Institute of Technology（数字孪生研究所、东部技术研究院）； University of Science and Technology of China（中国科学技术大学）； The Hong Kong University of Science and Technology（香港科技大学）

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

2510.12789 2025-10-15 cs.CV cs.AI cs.LG 62%

UniFusion: Vision-Language Model as Unified Encoder in Image Generation

Kevin Li, Manuel Brack, Sudeep Katakol, Hareesh Ravi, Ajinkya Kale

机构 * Adobe Applied Research（Adobe应用研究）

专题命中其他推理：reasoning(abstract)；分类 cs.AI、cs.LG

Comments Project page at https://thekevinli.github.io/unifusion/

2510.12699 2025-10-15 cs.CL cs.AI 62%

Generation Space Size: Understanding and Calibrating Open-Endedness of LLM Generations

Sunny Yu, Ahmad Jabbar, Robert Hawkins, Dan Jurafsky, Myra Cheng

机构 * Department of Computer Science, Stanford University（计算机科学系，斯坦福大学）； Department of Linguistics, Stanford University（语言学系，斯坦福大学）

专题命中其他推理：reasoning(abstract)；分类 cs.CL、cs.AI

2510.12245 2025-10-15 cs.LG cs.AI 62%

MoRA: On-the-fly Molecule-aware Low-Rank Adaptation Framework for LLM-based Multi-Modal Molecular Assistant

Tao Yin, Xiaohong Zhang, Jiacheng Zhang, Li Huang, Zhibin Zhang, Yuansong Zeng, Jin Xie, Meng Yan

专题命中其他推理：reasoning(abstract)；分类 cs.AI、cs.LG

1. 其他推理 5813 篇

Revealing Multimodal Causality with Large Language Models

Fine-Tuned Language Models for Domain-Specific Summarization and Tagging

Dingtalk DeepResearch: A Unified Multi Agent Framework for Adaptive Intelligence in Enterprise Environments

Mechanism and Emergence of Stacked Attention Heads in Multi-Layer Transformers

ViPER: Empowering the Self-Evolution of Visual Perception Abilities in Vision-Language Model

OmniVinci: Enhancing Architecture and Data for Omni-Modal Understanding LLM

AutoStreamPipe: LLM Assisted Automatic Generation of Data Stream Processing Pipelines

Can Language Models Compose Skills In-Context?

Critical Insights into Leading Conversational AI Models

Do Stop Me Now: Detecting Boilerplate Responses with a Single Iteration

Backward-Friendly Optimization: Training Large Language Models with Approximate Gradients under Memory Constraints

Aligning LLMs for Multilingual Consistency in Enterprise Applications

Language Model Guided Reinforcement Learning in Quantitative Trading

MOSAIC: Modeling Social AI for Content Dissemination and Regulation in Multi-Agent Simulations

A Frustratingly Simple Yet Highly Effective Attack Baseline: Over 90% Success Rate Against the Strong Black-box Models of GPT-4.5/4o/o1

A Hierarchical Error Framework for Reliable Automated Coding in Communication Research: Applications to Health and Political Communication

Empirical Evidence for Alignment Faking in a Small LLM and Prompt-Based Mitigation Techniques

LLM-Explorer: A Plug-in Reinforcement Learning Policy Exploration Enhancement Driven by Large Language Models

Plural Voices, Single Agent: Towards Inclusive AI in Multi-User Domestic Spaces

From Retrieval to Generation: Unifying External and Parametric Knowledge for Medical Question Answering

The Narcissus Hypothesis: Descending to the Rung of Illusion

UniRL-Zero: Reinforcement Learning on Unified Models with Joint Language Model and Diffusion Model Experts

Outraged AI: Large language models prioritise emotion over cost in fairness enforcement

UFT: Unifying Supervised and Reinforcement Fine-Tuning

What Limits Agentic Systems Efficiency?

Efficient Parallel Samplers for Recurrent-Depth Models and Their Connection to Diffusion Language Models

Informed Routing in LLMs: Smarter Token-Level Computation for Faster Inference

UniFusion: Vision-Language Model as Unified Encoder in Image Generation

Generation Space Size: Understanding and Calibrating Open-Endedness of LLM Generations

MoRA: On-the-fly Molecule-aware Low-Rank Adaptation Framework for LLM-based Multi-Modal Molecular Assistant