视觉大模型 / VLM - arXivDaily 专题

2511.20218 2025-11-26 cs.CV 70%

Text-guided Controllable Diffusion for Realistic Camouflage Images Generation

基于文本引导的可控扩散生成逼真伪装图像

Yuhang Qian, Haiyan Chen, Wentong Li, Ningzhong Liu, Jie Qin

专题命中 VLM训练与架构：VLM(abstract)；visual language model(abstract)；分类 cs.CV

AI总结本文提出CT-CIG方法，通过文本引导和可控扩散生成逼真且逻辑合理的伪装图像，利用VLM和FIRM模块提升伪装图像质量。

Comments Accepted by AAAI 2026

2511.19647 2025-11-26 cs.RO cs.AI 70%

Robot-Powered Data Flywheels: Deploying Robots in the Wild for Continual Data Collection and Foundation Model Adaptation

机器人驱动的数据飞轮：在真实世界中部署机器人以实现持续的数据收集和基础模型适应

Jennifer Grannen, Michelle Pan, Kenneth Llontop, Cherie Ho, Mark Zolotas, Jeannette Bohg, Dorsa Sadigh

机构 * Stanford University（斯坦福大学）； Toyota Research Institute（丰田研究机构）

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.AI

AI总结机器人驱动的数据飞轮通过部署机器人收集真实世界数据，提升基础模型在复杂环境中的适应能力，减少人工干预并提高多语言OCR性能。

2511.18875 2025-11-25 cs.CV cs.MM 70%

Parallel Vision Token Scheduling for Fast and Accurate Multimodal LMMs Inference

并行视觉令牌调度用于快速准确的多模态大语言模型推理

Wengyi Zhan, Mingbao Lin, Zhihang Lin, Rongrong Ji

机构 * Xiamen University（厦门大学）； Rakuten Asia Pte. Ltd.（拉面亚洲有限公司）； Xiamen University, China（厦门大学）； Shanghai Innovation Institute, China（上海创新研究院）

专题命中 VLM训练与架构：multimodal large language model(abstract)；MLLM(abstract)；分类 cs.CV

AI总结 ParVTS通过并行调度视觉令牌，高效减少多模态大语言模型推理的计算复杂度，实现高剪枝率与性能损失小的平衡。

2508.20840 2025-11-25 cs.RO cs.AI cs.MM 70%

Learning Primitive Embodied World Models: Towards Scalable Robotic Learning

学习原始具身世界模型：迈向可扩展的机器人学习

Qiao Sun, Liujia Yang, Wei Tang, Wei Huang, Kaixin Xu, Yongchao Chen, Mingyu Liu, Jiange Yang, Haoyi Zhu, Yating Wang, Tong He, Yilun Chen, Xili Dai, Nanyang Ye, Qinying Gu

机构 * Shanghai AI Lab（上海人工智能实验室）； Fudan（复旦大学）； SJTU（上海交通大学）； NJUST（南京理工大学）； THU（清华大学）； Harvard（哈佛大学）； ZJU（浙江大学）； NJU（南京大学）； USTC（中国科学技术大学）； Tongji（同济大学）； HKUST (GZ)（香港科技大学（广州））

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.AI

AI总结提出原始具身世界模型（PEWM）以解决具身数据稀疏和高维问题，通过短视界视频生成实现细粒度对齐、降低学习复杂度并提高数据效率。

2511.15923 2025-11-21 cs.CV 70%

RB-FT: Rationale-Bootstrapped Fine-Tuning for Video Classification

RB-FT：基于理由的视频分类微调

Meilong Xu, Di Fu, Jiaxing Zhang, Gong Yu, Jiayu Zheng, Xiaoling Hu, Dongdi Zhao, Feiyang Li, Chao Chen, Yong Cao

机构 * Stony Brook University（石溪大学）； ByteDance Inc.（字节跳动公司）； Harvard Medical School（哈佛医学院）

专题命中 VLM训练与架构：vision language model(abstract)；VLM(abstract)；分类 cs.CV

AI总结 RB-FT通过自动生成的理由提升视频分类性能，无需额外标注，有效提升模型对领域特定视频内容的理解能力。

Comments 11 pages, 2 figures

2511.14760 2025-11-19 cs.CV 70%

UniGen-1.5: Enhancing Image Generation and Editing through Reward Unification in Reinforcement Learning

Rui Tian, Mingfei Gao, Haiming Gang, Jiasen Lu, Zhe Gan, Yinfei Yang, Zuxuan Wu, Afshin Dehghan

机构 * Institute of Trustworthy Embodied AI, Fudan University（可信具身人工智能研究院，复旦大学）； Apple（苹果公司）

专题命中 VLM训练与架构：multimodal large language model(abstract)；MLLM(abstract)；分类 cs.CV

2506.23982 2025-11-19 cs.CV cs.RO 70%

StyleDrive: Towards Driving-Style Aware Benchmarking of End-To-End Autonomous Driving

Ruiyang Hao, Bowen Jing, Haibao Yu, Zaiqing Nie

机构 * AIR, Tsinghua University（空气动力学研究所，清华大学）； King’s College London（伦敦国王学院）； The University of Manchester（曼彻斯特大学）； The University of Hong Kong（香港大学）

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.CV

Comments 25 pages, 7 figures, 5 tables

2511.12917 2025-11-18 cs.CV 70%

Explore How to Inject Beneficial Noise in MLLMs

Ruishu Zhu, Sida Huang, Ziheng Jiao, Hongyuan Zhang

专题命中 VLM训练与架构：LLaVA(abstract)；multimodal large language model(abstract)；分类 cs.CV

Comments Accepted by AAAI 2026

2511.12525 2025-11-18 cs.CV 70%

MdaIF: Robust One-Stop Multi-Degradation-Aware Image Fusion with Language-Driven Semantics

Jing Li, Yifan Wang, Jiafeng Yan, Renlong Zhang, Bin Yang

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.CV

Comments 10 pages, 7 figures. Accepted by AAAI 2026

2511.09063 2025-11-17 cs.LG 70%

Human-Corrected Labels Learning: Enhancing Labels Quality via Human Correction of VLMs Discrepancies

Zhongnian Li, Lan Chen, Yixin Xu, Shi Xu, Xinzheng Xu

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.LG

Comments 11 pages, 3 figures

2511.10948 2025-11-17 cs.CV cs.HC 70%

DEFT-LLM: Disentangled Expert Feature Tuning for Micro-Expression Recognition

Ren Zhang, Huilai Li, Chao qi, Guoliang Xu, Tianyu Zhou, Wei wei, Jianqin Yin

机构 * College of Intelligent Engineering and Automation, Beijing University of Posts and Telecommunications（智能工程与自动化学院，北京邮电大学）

专题命中 VLM训练与架构：multimodal large language model(abstract)；MLLM(abstract)；分类 cs.CV

2507.02909 2025-11-14 cs.LG 70%

Fine-grained Token Allocation Via Operation Pruning for Efficient MLLMs

Aoming Liu, Reuben Tan, Boqing Gong, Bryan A. Plummer

机构 * Boston University（波士顿大学）； Microsoft Research（微软研究院）

专题命中 VLM训练与架构：LLaVA(abstract)；multimodal large language model(abstract)；分类 cs.LG

2511.07929 2025-11-12 cs.CV 70%

Federated CLIP for Resource-Efficient Heterogeneous Medical Image Classification

Yihang Wu, Ahmad Chaddad

专题命中 VLM训练与架构：vision language model(abstract)；VLM(abstract)；分类 cs.CV

Comments Accepted in AAAI 2026 Main track. Code is available at https://github.com/AIPMLab/FedMedCLIP

2507.10069 2025-11-12 cs.DC cs.LG 70%

ElasticMM: Efficient Multimodal LLMs Serving with Elastic Multimodal Parallelism

Zedong Liu, Shenggan Cheng, Guangming Tan, Yang You, Dingwen Tao

机构 * Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所）； University of Electronic Science and Technology of China（电子科技大学）； National University of Singapore（新加坡国立大学）

专题命中 VLM训练与架构：multimodal large language model(abstract)；MLLM(abstract)；分类 cs.LG

Comments Accepted at NeurIPS 2025 Oral (Thirty-Ninth Conference on Neural Information Processing Systems)

2511.06619 2025-11-11 cs.RO cs.AI 70%

How Do VLAs Effectively Inherit from VLMs?

Chuheng Zhang, Rushuai Yang, Xiaoyu Chen, Kaixin Wang, Li Zhao, Yi Chen, Jiang Bian

机构 * Microsoft Research（微软研究院）； The Hong Kong University of Science（香港科学大学）； Tsinghua University（清华大学）

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.AI

2510.15418 2025-11-10 cs.CL cs.AI 70%

Fine-Tuning MedGemma for Clinical Captioning to Enhance Multimodal RAG over Malaysia CPGs

Lee Qi Zun, Mohamad Zulhilmi Bin Abdul Halim, Goh Man Fye

专题命中 VLM训练与架构：vision-language model(abstract)；grounding(abstract)；分类 cs.AI

2506.11234 2025-11-07 cs.RO cs.CV 70%

Poutine: Vision-Language-Trajectory Pre-Training and Reinforcement Learning Post-Training Enable Robust End-to-End Autonomous Driving

Luke Rowe, Rodrigue de Schaetzen, Roger Girgis, Christopher Pal, Liam Paull

机构 * Mila - Quebec AI Institute（魁北克人工智能研究所）； Université de Montréal（蒙特利尔大学）； Polytechnique Montréal（蒙特利尔理工学院）； CIFAR AI Chair（CIFAR人工智能主席）

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.CV

2510.25067 2025-11-05 cs.CV 70%

DRIP: Dynamic patch Reduction via Interpretable Pooling

Yusen Peng, Sachin Kumar

专题命中 VLM训练与架构：vision-language model(abstract)；vision language model(abstract)；分类 cs.CV

Comments Need more refinement

2511.01791 2025-11-04 cs.RO cs.AI 70%

GenDexHand: Generative Simulation for Dexterous Hands

Feng Chen, Zhuxiu Xu, Tianzhe Chu, Xunzhe Zhou, Li Sun, Zewen Wu, Shenghua Gao, Zhongyu Li, Yanchao Yang, Yi Ma

机构 * The University of Hong Kong（香港大学）； Transcengram ； Shanghai Jiao Tong University（上海交通大学）； The Chinese University of Hong Kong（香港中文大学）

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.AI

2511.00076 2025-11-04 cs.LG 70%

Bridging Vision, Language, and Mathematics: Pictographic Character Reconstruction with Bézier Curves

Zihao Wan, Pau Tong Lin Xu, Fuwen Luo, Ziyue Wang, Peng Li, Yang Liu

机构 * Dept. of Comp. Sci. & Tech., Institute for AI, Tsinghua University, Beijing, China（计算机科学与技术系，人工智能研究院，清华大学，北京，中国）； Institute for AI Industry Research (AIR), Tsinghua University, Beijing, China（人工智能产业研究院（AIR），清华大学，北京，中国）

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.LG

2511.00041 2025-11-04 cs.RO cs.AI 70%

Endowing GPT-4 with a Humanoid Body: Building the Bridge Between Off-the-Shelf VLMs and the Physical World

Yingzhao Jian, Zhongan Wang, Yi Yang, Hehe Fan

机构 * College of Computer Science and Technology（计算机科学与技术学院）； Zhejiang University（浙江大学）

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.AI

2506.05696 2025-10-31 cs.CV 70%

MoralCLIP: Contrastive Alignment of Vision-and-Language Representations with Moral Foundations Theory

Ana Carolina Condez, Diogo Tavares, João Magalhães

机构 * NOVA LINCS, NOVA School of Science and Technology（NOVA LINCS，NOVA科学与技术学院）

专题命中 VLM训练与架构：vision-language model(abstract)；grounding(abstract)；分类 cs.CV

Comments Updated version: corresponds to the ACM MM '25 published paper and includes full appendix material

2510.24214 2025-10-29 cs.CV 70%

SCOPE: Saliency-Coverage Oriented Token Pruning for Efficient Multimodel LLMs

Jinhong Deng, Wen Li, Joey Tianyi Zhou, Yang He

机构 * University of Electronic Science and Technology of China（电子科技大学）； Shenzhen Institute for Advanced Study（深圳先进研究 institute）； CFAR, Agency for Science, Technology and Research (A*STAR)（科技研究局（A*STAR）认知与人工智能研究中心）； IHPC, Agency for Science, Technology and Research (A*STAR)（科技研究局（A*STAR）人工智能中心）

专题命中 VLM训练与架构：LLaVA(abstract)；multimodal large language model(abstract)；分类 cs.CV

Comments NeurIPS 2025

2510.23203 2025-10-28 cs.CV 70%

DecoDINO: 3D Human-Scene Contact Prediction with Semantic Classification

Lukas Bierling, Davide Pasero, Fleur Dolmans, Helia Ghasemi, Angelo Broere

机构 * University of Amsterdam（阿姆斯特丹大学）

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.CV

2510.17847 2025-10-22 cs.CV 70%

CoIDO: Efficient Data Selection for Visual Instruction Tuning via Coupled Importance-Diversity Optimization

Yichen Yan, Ming Zhong, Qi Zhu, Xiaoling Gu, Jinpeng Chen, Huan Li

机构 * The State Key Laboratory of Blockchain and Data Security（区块链与数据安全国家重点实验室）； Hangzhou High-Tech Zone (Binjiang) Institute of Blockchain and Data Security（杭州高新技术区（滨江）区块链与数据安全研究院）； Hangzhou Dianzi University（杭州电子科技大学）； School of Computer Science (National Pilot Software Engineering School), BUPT（计算机学院（国家级试点软件工程学院），北京邮电大学）

专题命中 VLM训练与架构：LLaVA(abstract)；multimodal large language model(abstract)；分类 cs.CV

Comments 22 pages, 8 figures, 39th Conference on Neural Information Processing Systems (NeurIPS 2025)

2510.17205 2025-10-21 cs.CV cs.CL 70%

$\mathcal{V}isi\mathcal{P}runer$: Decoding Discontinuous Cross-Modal Dynamics for Efficient Multimodal LLMs

Yingqi Fan, Anhao Zhao, Jinlan Fu, Junlong Tong, Hui Su, Yijie Pan, Wei Zhang, Xiaoyu Shen

机构 * Ningbo Key Laboratory of Spatial Intelligence and Digital Derivative, Institute of Digital Twin, EIT, Ningbo（宁波空间智能与数字衍生关键实验室，数字孪生研究院，EIT，宁波）； Shanghai Jiao Tong University（上海交通大学）； Hong Kong Polytechnic University（香港理工大学）； Meituan Inc.（美团公司）； National University of Singapore（新加坡国立大学）

专题命中 VLM训练与架构：LLaVA(abstract)；multimodal large language model(abstract)；分类 cs.CV

Comments EMNLP 2025 Main

2507.13061 2025-10-21 cs.CV 70%

Advancing Complex Wide-Area Scene Understanding with Hierarchical Coresets Selection

Jingyao Wang, Yiming Chen, Lingyu Si, Changwen Zheng

机构 * Institute of Software Chinese Academy of Sciences（中国科学院软件研究所）； University of the Chinese Academy of Sciences（中国科学院大学）； Beijing University of Technology（北京理工大学）

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.CV

Comments Accepted by ACMMM2025

2510.12548 2025-10-15 cs.CL cs.CV 70%

VISaGE: Understanding Visual Generics and Exceptions

Stella Frank, Emily Allaway

机构 * University of Copenhagen（哥本哈根大学）； University of Edinburgh（爱丁堡大学）

专题命中 VLM训练与架构：vision language model(abstract)；VLM(abstract)；分类 cs.CV

Comments EMNLP 2025

2407.09073 2025-10-14 cs.CV 70%

Open Vocabulary Multi-Label Video Classification

Rohit Gupta, Mamshad Nayeem Rizve, Jayakrishnan Unnikrishnan, Ashish Tawari, Son Tran, Mubarak Shah, Benjamin Yao, Trishul Chilimbi

机构 * Center for Research in Computer Vision, University of Central Florida（计算机视觉研究中心，中央佛罗里达大学）； Amazon（亚马逊）

专题命中 VLM训练与架构：vision-language model(abstract)；VLM(abstract)；分类 cs.CV

Comments Accepted at ECCV 2024

2510.09699 2025-10-14 cs.CR cs.AI 70%

VisualDAN: Exposing Vulnerabilities in VLMs with Visual-Driven DAN Commands

Aofan Liu, Lulu Tang

机构 * Beijing Academy of Artificial Intelligence（北京人工智能研究院）

专题命中 VLM训练与架构：vision-language model(abstract)；LLaVA(abstract)；分类 cs.AI

1. VLM训练与架构 5048 篇

Text-guided Controllable Diffusion for Realistic Camouflage Images Generation

Robot-Powered Data Flywheels: Deploying Robots in the Wild for Continual Data Collection and Foundation Model Adaptation

Parallel Vision Token Scheduling for Fast and Accurate Multimodal LMMs Inference

Learning Primitive Embodied World Models: Towards Scalable Robotic Learning

RB-FT: Rationale-Bootstrapped Fine-Tuning for Video Classification

UniGen-1.5: Enhancing Image Generation and Editing through Reward Unification in Reinforcement Learning

StyleDrive: Towards Driving-Style Aware Benchmarking of End-To-End Autonomous Driving

Explore How to Inject Beneficial Noise in MLLMs

MdaIF: Robust One-Stop Multi-Degradation-Aware Image Fusion with Language-Driven Semantics

Human-Corrected Labels Learning: Enhancing Labels Quality via Human Correction of VLMs Discrepancies

DEFT-LLM: Disentangled Expert Feature Tuning for Micro-Expression Recognition

Fine-grained Token Allocation Via Operation Pruning for Efficient MLLMs

Federated CLIP for Resource-Efficient Heterogeneous Medical Image Classification

ElasticMM: Efficient Multimodal LLMs Serving with Elastic Multimodal Parallelism

How Do VLAs Effectively Inherit from VLMs?

Fine-Tuning MedGemma for Clinical Captioning to Enhance Multimodal RAG over Malaysia CPGs

Poutine: Vision-Language-Trajectory Pre-Training and Reinforcement Learning Post-Training Enable Robust End-to-End Autonomous Driving

DRIP: Dynamic patch Reduction via Interpretable Pooling

GenDexHand: Generative Simulation for Dexterous Hands

Bridging Vision, Language, and Mathematics: Pictographic Character Reconstruction with Bézier Curves

Endowing GPT-4 with a Humanoid Body: Building the Bridge Between Off-the-Shelf VLMs and the Physical World

MoralCLIP: Contrastive Alignment of Vision-and-Language Representations with Moral Foundations Theory

SCOPE: Saliency-Coverage Oriented Token Pruning for Efficient Multimodel LLMs

DecoDINO: 3D Human-Scene Contact Prediction with Semantic Classification

CoIDO: Efficient Data Selection for Visual Instruction Tuning via Coupled Importance-Diversity Optimization

$\mathcal{V}isi\mathcal{P}runer$: Decoding Discontinuous Cross-Modal Dynamics for Efficient Multimodal LLMs

Advancing Complex Wide-Area Scene Understanding with Hierarchical Coresets Selection

VISaGE: Understanding Visual Generics and Exceptions

Open Vocabulary Multi-Label Video Classification

VisualDAN: Exposing Vulnerabilities in VLMs with Visual-Driven DAN Commands