视觉大模型 / VLM - arXivDaily 专题

2510.06679 2025-10-09 cs.CV 57%

DreamOmni2: Multimodal Instruction-based Editing and Generation

Bin Xia, Bohao Peng, Yuechen Zhang, Junjia Huang, Jiyang Liu, Jingyao Li, Haoru Tan, Sitong Wu, Chengyao Wang, Yitong Wang, Xinglong Wu, Bei Yu, Jiaya Jia

机构 * CUHK（香港中文大学）； HKUST（香港科技大学）； HKU（香港大学）； ByteDance Inc（字节跳动公司）

专题命中 VLM训练与架构：VLM(abstract)；分类 cs.CV

2501.19252 2025-10-08 cs.CV 57%

Inference-Time Text-to-Video Alignment with Diffusion Latent Beam Search

Yuta Oshima, Masahiro Suzuki, Yutaka Matsuo, Hiroki Furuta

机构 * The University of Tokyo（东京大学）； Google DeepMind（谷歌DeepMind）

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

Comments Accepted to NeurIPS2025. Website: https://sites.google.com/view/t2v-dlbs and Code: https://github.com/shim0114/T2V-Diffusion-Search

2509.22229 2025-10-07 cs.CV 57%

A Tale of Two Experts: Cooperative Learning for Source-Free Unsupervised Domain Adaptation

Jiaping Yu, Muli Yang, Jiapeng Ji, Jiexi Yan, Cheng Deng

机构 * School of Electronic Engineering（电子工程学院）； Xidian University（西安电子科技大学）； School of Computer Science（计算机科学学院）

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

2506.03591 2025-10-07 cs.CV 57%

Resolving Task Objective Conflicts in Unified Model via Task-Aware Mixture-of-Experts

Jiaxing Zhang, Hao Tang

机构 * Sichuan University（四川大学）； Peking University（北京大学）

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

2510.02778 2025-10-06 cs.CV 57%

AdaRD-key: Adaptive Relevance-Diversity Keyframe Sampling for Long-form Video understanding

Xian Zhang, Zexi Wu, Zinuo Li, Hongming Xu, Luqi Gong, Farid Boussaid, Naoufel Werghi, Mohammed Bennamoun

机构 * The University of Western Australia（西澳大学）； Dalian University of Technology（大连理工大学）； Khalifa University（卡利夫大学）； Zhejiang Lab（浙江实验室）

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

2509.00320 2025-10-03 cs.CV 57%

TrimTokenator: Towards Adaptive Visual Token Pruning for Large Multimodal Models

Hao Zhang, Mengsi Lyu, Chenrui He, Yulong Ao, Yonghua Lin

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

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

Comments 15 pages

2502.01051 2025-10-03 cs.CV 57%

Diffusion Model as a Noise-Aware Latent Reward Model for Step-Level Preference Optimization

Tao Zhang, Cheng Da, Kun Ding, Huan Yang, Kun Jin, Yan Li, Tingting Gao, Di Zhang, Shiming Xiang, Chunhong Pan

机构 * MAIS, CASIA（中国科学院自动化研究所MAIS实验室）； Kuaishou Technology（快手科技）； School of Artificial Intelligence, UCAS（中国科学院大学人工智能学院）

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

Comments NeurIPS 2025

2510.00993 2025-10-02 cs.CV 57%

Visual Self-Refinement for Autoregressive Models

Jiamian Wang, Ziqi Zhou, Chaithanya Kumar Mummadi, Sohail Dianat, Majid Rabbani, Raghuveer Rao, Chen Qiu, Zhiqiang Tao

机构 * Rochester Institute of Technology（罗切斯特理工大学）； Bosch Research（博世研究）； DEVCOM Army Research Laboratory（美国陆军研究实验室）

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

Comments Accepted by EMNLP2025

2510.00618 2025-10-02 cs.CV 57%

Robust Context-Aware Object Recognition

Klara Janouskova, Cristian Gavrus, Jiri Matas

机构 * Visual Recognition Group, Czech Technical University in Prague（捷克技术大学布拉格分校视觉识别组）

专题命中 VLM训练与架构：VLM(abstract)；分类 cs.CV

2509.24200 2025-10-01 cs.CV 57%

UniVid: The Open-Source Unified Video Model

Jiabin Luo, Junhui Lin, Zeyu Zhang, Biao Wu, Meng Fang, Ling Chen, Hao Tang

机构 * Peking University（北京大学）； AI Geeks ； Australian Artificial Intelligence Institute（澳大利亚人工智能研究所）

专题命中 VLM训练与架构：MLLM(abstract)；分类 cs.CV

2503.16257 2025-09-30 cs.CV 57%

Plug-and-Play 1.x-Bit KV Cache Quantization for Video Large Language Models

Keda Tao, Haoxuan You, Yang Sui, Can Qin, Huan Wang

机构 * Zhejiang University（浙江大学）； Westlake University（西湖大学）； Columbia University（哥伦比亚大学）； Rice University（Rice大学）； Salesforce AI Research（Salesforce人工智能研究）

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

Comments 12 pages

2509.23760 2025-09-30 cs.CV 57%

UniAlignment: Semantic Alignment for Unified Image Generation, Understanding, Manipulation and Perception

Xinyang Song, Libin Wang, Weining Wang, Shaozhen Liu, Dandan Zheng, Jingdong Chen, Qi Li, Zhenan Sun

机构 * Ant Group（蚂蚁集团）

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

2509.22570 2025-09-29 cs.AI 57%

UniMIC: Token-Based Multimodal Interactive Coding for Human-AI Collaboration

Qi Mao, Tinghan Yang, Jiahao Li, Bin Li, Libiao Jin, Yan Lu

机构 * State Key Laboratory of Media Convergence and Communication（媒体融合与传播国家重点实验室）； Communication University of China（中国传媒大学）； Microsoft Research Asia（微软亚洲研究院）

专题命中 VLM训练与架构：visual question answering(abstract)；分类 cs.AI

2509.21287 2025-09-26 cs.CL cs.AI 57%

DisCoCLIP: A Distributional Compositional Tensor Network Encoder for Vision-Language Understanding

Kin Ian Lo, Hala Hawashin, Mina Abbaszadeh, Tilen Limback-Stokin, Hadi Wazni, Mehrnoosh Sadrzadeh

机构 * University College London（伦敦大学学院）

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

2502.12148 2025-09-26 cs.CV 57%

HermesFlow: Seamlessly Closing the Gap in Multimodal Understanding and Generation

Ling Yang, Xinchen Zhang, Ye Tian, Chenming Shang, Minghao Xu, Wentao Zhang, Bin Cui

机构 * Peking University（北京大学）； Tsinghua University（清华大学）； Mila - Québec AI Institute（魁北克人工智能研究所）

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

Comments NeurIPS 2025. Code: https://github.com/Gen-Verse/HermesFlow

2509.19018 2025-09-24 cs.LG 57%

OmniBridge: Unified Multimodal Understanding, Generation, and Retrieval via Latent Space Alignment

Teng Xiao, Zuchao Li, Lefei Zhang

机构 * School of Computer Science, Wuhan University（武汉大学计算机学院）； School of Artificial Intelligence, Wuhan University（武汉大学人工智能学院）

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

2509.18778 2025-09-24 cs.RO cs.AI 57%

VGGT-DP: Generalizable Robot Control via Vision Foundation Models

Shijia Ge, Yinxin Zhang, Shuzhao Xie, Weixiang Zhang, Mingcai Zhou, Zhi Wang

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

Comments submitted to AAAI 2026

2503.04130 2025-09-24 cs.CV 57%

STORM: Token-Efficient Long Video Understanding for Multimodal LLMs

Jindong Jiang, Xiuyu Li, Zhijian Liu, Muyang Li, Guo Chen, Zhiqi Li, De-An Huang, Guilin Liu, Zhiding Yu, Kurt Keutzer, Sungjin Ahn, Jan Kautz, Hongxu Yin, Yao Lu, Song Han, Wonmin Byeon

机构 * NVIDIA（英伟达）； Rutgers University（罗格斯大学）； UC Berkeley（加州大学伯克利分校）； MIT（麻省理工学院）； Nanjing University（南京大学）； KAIST（韩国科学技术院）

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

2509.17065 2025-09-23 cs.CV 57%

CardiacCLIP: Video-based CLIP Adaptation for LVEF Prediction in a Few-shot Manner

Yao Du, Jiarong Guo, Xiaomeng Li

机构 * Department of Electronic and Computer Engineering, The Hong Kong University of Science and Technology（电子与计算机工程系，香港科学与技术大学）

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

Comments Accepted by MICCAI 2025

2509.16768 2025-09-23 cs.CV 57%

MMPart: Harnessing Multi-Modal Large Language Models for Part-Aware 3D Generation

Omid Bonakdar, Nasser Mozayani

机构 * School of Computer engineering, Iran university of Science and Technology（计算机工程学院，伊朗科学技术大学）

专题命中 VLM训练与架构：VLM(abstract)；分类 cs.CV

2509.15277 2025-09-22 cs.MM cs.LG 57%

Copycat vs. Original: Multi-modal Pretraining and Variable Importance in Box-office Prediction

Qin Chao, Eunsoo Kim, Boyang Li

机构 * College of Computing and Data Science, Nanyang Technological University, Singapore（计算与数据科学学院，南洋理工大学，新加坡）； Business School, University of Seoul, Republic of Korea（商学学院，首尔大学，韩国）； Alibaba Group and the Alibaba-NTU Joint Research Institute, Singapore（阿里巴巴集团及阿里巴巴-南洋理工大学联合研究机构，新加坡）

专题命中 VLM训练与架构：grounding(abstract)；分类 cs.LG

2411.14982 2025-09-19 cs.CV cs.CL 57%

Large Multi-modal Models Can Interpret Features in Large Multi-modal Models

Kaichen Zhang, Yifei Shen, Bo Li, Ziwei Liu

机构 * S-Lab, NTU, Singapore（新加坡国立大学S实验室）； LMMs-Lab Team（多模态模型实验室团队）； Microsoft Research Asia（微软亚洲研究院）

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

2509.14060 2025-09-18 cs.CV 57%

VSE-MOT: Multi-Object Tracking in Low-Quality Video Scenes Guided by Visual Semantic Enhancement

Jun Du, Weiwei Xing, Ming Li, Fei Richard Yu

机构 * School of Software Engineering（软件工程学院）； Guangdong Laboratory of Artificial Intelligence and Digital Economy（人工智能与数字经济广东实验室）

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

2509.13760 2025-09-18 cs.CV 57%

Iterative Prompt Refinement for Safer Text-to-Image Generation

Jinwoo Jeon, JunHyeok Oh, Hayeong Lee, Byung-Jun Lee

机构 * Korea University（韩国大学）

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

2505.21233 2025-09-18 cs.CV 57%

CROP: Contextual Region-Oriented Visual Token Pruning

Jiawei Guo, Feifei Zhai, Pu Jian, Qianrun Wei, Yu Zhou

机构 * State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, CAS, Beijing, China（多模态人工智能系统国家重点实验室，自动化研究所，中国科学院，北京，中国）； School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China（人工智能学院，中国科学院大学，北京，中国）； Fanyu AI Laboratory, Zhongke Fanyu Technology Co., Ltd, Beijing, China（凡语AI实验室，中科创始人技术有限公司，北京，中国）

专题命中 VLM训练与架构：VLM(abstract)；分类 cs.CV

Comments EMNLP2025 Main

2509.08858 2025-09-15 cs.CY cs.LG 57%

Decentralising LLM Alignment: A Case for Context, Pluralism, and Participation

Oriane Peter, Kate Devlin

专题命中 VLM训练与架构：grounding(abstract)；分类 cs.LG

Comments Accepted at AIES 2025

2508.10899 2025-09-15 cs.LG 57%

A Dataset for Distilling Knowledge Priors from Literature for Therapeutic Design

Haydn Thomas Jones, Natalie Maus, Josh Magnus Ludan, Maggie Ziyu Huan, Jiaming Liang, Marcelo Der Torossian Torres, Jiatao Liang, Zachary Ives, Yoseph Barash, Cesar de la Fuente-Nunez, Jacob R. Gardner, Mark Yatskar

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

2509.08940 2025-09-12 cs.CV 57%

Discovering Divergent Representations between Text-to-Image Models

Lisa Dunlap, Joseph E. Gonzalez, Trevor Darrell, Fabian Caba Heilbron, Josef Sivic, Bryan Russell

专题命中 VLM训练与架构：VLM(abstract)；分类 cs.CV

Comments Accepted to ICCV 2025. Code available at https://github.com/adobe-research/CompCon

2504.02438 2025-09-11 cs.CL cs.AI 57%

Scaling Video-Language Models to 10K Frames via Hierarchical Differential Distillation

Chuanqi Cheng, Jian Guan, Wei Wu, Rui Yan

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）； School of Artificial Intelligence, Wuhan University（武汉大学人工智能学院）

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

Comments Accepted by ICML 2025

2212.03221 2025-09-09 cs.CV eess.IV 57%

ADIR: Adaptive Diffusion for Image Reconstruction

Shady Abu-Hussein, Tom Tirer, Raja Giryes

机构 * Tel Aviv University（特拉维夫大学）； Bar Ilan University（巴伊兰大学）

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

Comments Project page https://shadyabh.github.io/ADIR/

Journal ref BMVC 2025

1. VLM训练与架构 5058 篇

DreamOmni2: Multimodal Instruction-based Editing and Generation

Inference-Time Text-to-Video Alignment with Diffusion Latent Beam Search

A Tale of Two Experts: Cooperative Learning for Source-Free Unsupervised Domain Adaptation

Resolving Task Objective Conflicts in Unified Model via Task-Aware Mixture-of-Experts

AdaRD-key: Adaptive Relevance-Diversity Keyframe Sampling for Long-form Video understanding

TrimTokenator: Towards Adaptive Visual Token Pruning for Large Multimodal Models

Diffusion Model as a Noise-Aware Latent Reward Model for Step-Level Preference Optimization

Visual Self-Refinement for Autoregressive Models

Robust Context-Aware Object Recognition

UniVid: The Open-Source Unified Video Model

Plug-and-Play 1.x-Bit KV Cache Quantization for Video Large Language Models

UniAlignment: Semantic Alignment for Unified Image Generation, Understanding, Manipulation and Perception

UniMIC: Token-Based Multimodal Interactive Coding for Human-AI Collaboration

DisCoCLIP: A Distributional Compositional Tensor Network Encoder for Vision-Language Understanding

HermesFlow: Seamlessly Closing the Gap in Multimodal Understanding and Generation

OmniBridge: Unified Multimodal Understanding, Generation, and Retrieval via Latent Space Alignment

VGGT-DP: Generalizable Robot Control via Vision Foundation Models

STORM: Token-Efficient Long Video Understanding for Multimodal LLMs

CardiacCLIP: Video-based CLIP Adaptation for LVEF Prediction in a Few-shot Manner

MMPart: Harnessing Multi-Modal Large Language Models for Part-Aware 3D Generation

Copycat vs. Original: Multi-modal Pretraining and Variable Importance in Box-office Prediction

Large Multi-modal Models Can Interpret Features in Large Multi-modal Models

VSE-MOT: Multi-Object Tracking in Low-Quality Video Scenes Guided by Visual Semantic Enhancement

Iterative Prompt Refinement for Safer Text-to-Image Generation

CROP: Contextual Region-Oriented Visual Token Pruning

Decentralising LLM Alignment: A Case for Context, Pluralism, and Participation

A Dataset for Distilling Knowledge Priors from Literature for Therapeutic Design

Discovering Divergent Representations between Text-to-Image Models

Scaling Video-Language Models to 10K Frames via Hierarchical Differential Distillation

ADIR: Adaptive Diffusion for Image Reconstruction