视觉大模型 / VLM

Zero-shot visual reasoning through probabilistic analogical mapping

Taylor W. Webb, Shuhao Fu, Trevor Bihl, Keith J. Holyoak, Hongjing Lu

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV、cs.AI

A Benchmark for Compositional Visual Reasoning

Aimen Zerroug, Mohit Vaishnav, Julien Colin, Sebastian Musslick, Thomas Serre

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV、cs.AI

Understanding the computational demands underlying visual reasoning

Mohit Vaishnav, Remi Cadene, Andrea Alamia, Drew Linsley, Rufin VanRullen, Thomas Serre

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV、cs.AI

Comments 26 pages, 16 figures

Journal ref Neural Computation, 2022

Grounding Complex Navigational Instructions Using Scene Graphs

Michiel de Jong, Satyapriya Krishna, Anuva Agarwal

专题命中 视觉推理 ：grounding(title)；visual question answering(abstract)；分类 cs.CV、cs.LG

Comments arXiv admin note: text overlap with arXiv:1706.07230 by other authors

Comparing Visual Reasoning in Humans and AI

Shravan Murlidaran, William Yang Wang, Miguel P. Eckstein

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV、cs.AI

Dynamic Language Binding in Relational Visual Reasoning

Thao Minh Le, Vuong Le, Svetha Venkatesh, Truyen Tran

专题命中 视觉推理 ：visual reasoning(title)；visual question answering(abstract)；分类 cs.CV、cs.LG

Comments Early version accepted by IJCAI20, Code available at https://github.com/thaolmk54/LOGNet-VQA

Meta Module Network for Compositional Visual Reasoning

Wenhu Chen, Zhe Gan, Linjie Li, Yu Cheng, William Wang, Jingjing Liu

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV、cs.LG

Comments Accepted to WACV 21 (Oral)

Few-shot Visual Reasoning with Meta-analogical Contrastive Learning

Youngsung Kim, Jinwoo Shin, Eunho Yang, Sung Ju Hwang

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.AI、cs.LG

Modeling Gestalt Visual Reasoning on the Raven's Progressive Matrices Intelligence Test Using Generative Image Inpainting Techniques

Tianyu Hua, Maithilee Kunda

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV、cs.LG

Few-Shot Abstract Visual Reasoning With Spectral Features

Tanner Bohn, Yining Hu, Charles X. Ling

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV、cs.LG

Comments 11 pages, 3 figures

CLEVR-Ref+: Diagnosing Visual Reasoning with Referring Expressions

Runtao Liu, Chenxi Liu, Yutong Bai, Alan Yuille

专题命中 视觉推理 ：visual reasoning(title)；visual question answering(abstract)；分类 cs.CV、cs.LG

Comments To appear in CVPR 2019. All data and code concerning CLEVR-Ref+ and IEP-Ref have been released at https://cs.jhu.edu/~cxliu/2019/clevr-ref+

Object Ordering with Bidirectional Matchings for Visual Reasoning

Hao Tan, Mohit Bansal

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV、cs.AI

Comments NAACL 2018 (8 pages; added pointer-ordering examples)

FiLM: Visual Reasoning with a General Conditioning Layer

Ethan Perez, Florian Strub, Harm de Vries, Vincent Dumoulin, Aaron Courville

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV、cs.AI

Comments AAAI 2018. Code available at http://github.com/ethanjperez/film . Extends arXiv:1707.03017

Learning Visual Reasoning Without Strong Priors

Ethan Perez, Harm de Vries, Florian Strub, Vincent Dumoulin, Aaron Courville

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV、cs.AI

Comments Full AAAI 2018 paper is at arXiv:1709.07871. Presented at ICML 2017's Machine Learning in Speech and Language Processing Workshop. Code is at http://github.com/ethanjperez/film

Inferring and Executing Programs for Visual Reasoning

Justin Johnson, Bharath Hariharan, Laurens van der Maaten, Judy Hoffman, Li Fei-Fei, C. Lawrence Zitnick, Ross Girshick

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV、cs.LG

Building Reasonable Inference for Vision-Language Models in Blind Image Quality Assessment

构建合理的视觉-语言模型推理以实现盲图像质量评估

Yuan Li, Zitang Sun, Yen-ju Chen, Shin'ya Nishida

机构 * Graduate School of Informatics, Kyoto University（京都大学信息学研究生院） 纠错

专题命中 视觉推理 ：vision-language model(title,journal_ref)；VLM(abstract)；分类 cs.CV

AI总结 本文提出一种两阶段调优方法，通过分离视觉感知与质量推断，提升视觉-语言模型在盲图像质量评估中的推理稳定性与可靠性。

Comments Accepted to the ICONIP (International Conference on Neural Information Processing), 2025

Journal ref Building Reasonable Inference for Vision-Language Models in Blind Image Quality Assessment. In: Taniguchi, T., et al. Neural Information Processing. ICONIP 2025. Lecture Notes in Computer Science, vol 16310. Springer, Singapore

Perspective-Aware Reasoning in Vision-Language Models via Mental Imagery Simulation

Phillip Y. Lee, Jihyeon Je, Chanho Park, Mikaela Angelina Uy, Leonidas Guibas, Minhyuk Sung

专题命中 视觉推理 ：vision-language model(title,abstract)；分类 cs.CV；VLM(comments)

Comments Project Page: https://apc-vlm.github.io/

Can Vision Language Models Learn from Visual Demonstrations of Ambiguous Spatial Reasoning?

Bowen Zhao, Leo Parker Dirac, Paulina Varshavskaya

专题命中 视觉推理 ：vision language model(title)；vision-language model(abstract)；分类 cs.CV；VLM(comments)

Comments 13 pages, 4 figures. Code released at https://github.com/groundlight/vlm-visual-demonstrations

Deep Reason: A Strong Baseline for Real-World Visual Reasoning

Chenfei Wu, Yanzhao Zhou, Gen Li, Nan Duan, Duyu Tang, Xiaojie Wang

专题命中 视觉推理 ：visual reasoning(title,abstract)；分类 cs.CV；visual question answering(comments)

Comments CVPR 2019 Visual Question Answering and Dialog Workshop

CP-MoE: Consistency-Preserving Mixture-of-Experts for Continual Learning

CP-MoE：一致性保留的混合专家用于持续学习

Yang Liu, Toan Nguyen, Flora D. Salim

机构 * School of Computer Science and Engineering University of New South Wales（计算机科学与工程学院 新南威尔士大学） 纠错

专题命中 视觉推理 ：VLM(abstract,abstract_cn)；visual reasoning(abstract)；分类 cs.CV、cs.AI、cs.LG

AI总结 本文提出CP-MoE，一种基于瞬时专家的持续学习框架，通过一致性保留的路由偏置和瞬时专家引导的正则化机制，减少参数干扰和遗忘，同时保留跨任务知识转移。

Comments Accepted at CoLLAs 2026

MOON3.0: Reasoning-aware Multimodal Representation Learning for E-commerce Product Understanding

MOON3.0: 基于推理的多模态表示学习用于电商产品理解

Junxian Wu, Chenghan Fu, Zhanheng Nie, Daoze Zhang, Bowen Wan, Wanxian Guan, Chuan Yu, Jian Xu, Bo Zheng

机构 * Alibaba Group（阿里巴巴集团） 纠错

专题命中 视觉推理 ：MLLM(abstract,abstract_cn)；multimodal large language model(abstract)；分类 cs.CV、cs.AI、cs.LG

AI总结 MOON3.0通过多头模态融合、联合对比与强化学习框架及细粒度残差增强模块，提升电商产品细粒度属性建模能力，并在大规模多模态电商基准MBe3.0上实现零样本最优性能。

Comments Accepted by the 34th ACM International Conference on Multimedia (ACM MM), 2026. 10 pages, 6 figures

D-VLC: Decentralized Vision-Language Collaboration for Heterogeneous Embodied Multi-Robot Systems in Unknown Environments

D-VLC：面向未知环境中异构具身多机器人系统的去中心化视觉-语言协作框架

Yuan Zhou, Ruitong Lin, Shen Wang, Weiqi Gai, Mo Zhu, Xin Zhou, Yuze Wu, Fei Gao

专题命中 视觉推理 ：VLM(summary_cn,abstract)

AI总结 本文提出D-VLC框架，结合去中心化异步推理等技术，让通用VLM生成机器人特定动作，多场景实验显示其任务成功率超70%，完成时间较几何贪心基线最多缩短55.8%。

From Pixels to Concepts: Growing Rich 3D Semantic Scene Graph Forests utilizing Foundation Models

从像素到概念：利用基础模型构建丰富的3D语义场景图森林

David Oberacker, Meike Deitersen, Niklas Spielbauer, Tristan Schnell, Georg Heppner, Arne Roennau

机构 * FZI Research Center for Information Technology（FZI信息技术研究中心） ； Machine Intelligence and Robotics Lab (MaiRo), Karlsruhe Institute for Technology (KIT)（卡尔斯鲁厄理工学院机器智能与机器人实验室） 纠错

专题命中 视觉推理 ：VLM(summary_cn,abstract)

AI总结 提出利用基础模型构建具有开放语义关系的3D场景图森林，通过VLM和LLM推理抽象概念节点与关系，提升机器人场景理解与任务执行能力。

Comments To be published in the Proceedings of the IEEE/RSJ International Conference on Intelligent Robots & Systems (IEEE IROS 2026)

OneCanvas: 3D Scene Understanding via Panoramic Reprojection

OneCanvas: 通过全景重投影实现3D场景理解

Bartłomiej Baranowski, Dave Zhenyu Chen, Matthias Nießner

机构 * Technical University of Munich（慕尼黑技术大学） ； Huawei（华为） 纠错

专题命中 视觉推理 ：VLM(abstract,abstract_cn)；vision-language model(abstract)；分类 cs.CV、cs.AI、cs.LG

AI总结 提出OneCanvas方法，将多视图补丁特征聚合到全景画布上，利用深度和相机位姿进行重投影，无需复杂几何编码器或大量训练，在SQA3D等基准上达到最先进精度。

Comments Project page: https://baranowskibrt.github.io/onecanvas/

Gen-VCoT: Generative Visual Chain-of-Thought Reasoning via Diffusion-Based RGB Intermediate Representations

Gen-VCoT: 基于扩散的RGB中间表示的生成式视觉思维链推理

Zhiqiang Zhou, Junliang Dai, Xu ling

机构 * Hunan Chemical Industry Vocational and Technical College（湖南化工职业技术学院） 纠错

专题命中 视觉推理 ：visual reasoning(abstract)；grounding(abstract)；multimodal large language model(abstract)；分类 cs.CV、cs.AI、cs.LG

AI总结 提出Gen-VCoT框架，利用专家视觉模型生成RGB图像作为推理中间步骤，通过自适应路由器选择推理深度，在空间和深度问题上分别提升25%和50%，但简单事实查询性能下降，表明最优表示依赖于任务。

Comments 12 pages, 5 figures

CycliST: A Video Language Model Benchmark for Reasoning on Cyclical State Transitions

CycliST：用于循环状态转换推理的视频语言模型基准

Simon Kohaut, Daniel Ochs, Shun Zhang, Benedict Flade, Julian Eggert, Kristian Kersting, Devendra Singh Dhami

机构 * Artificial Intelligence and Machine Learning Lab, TU Darmstadt（人工智能与机器学习实验室，图腾斯达特技术大学） ； Konrad Zuse School of Excellence in Learning and Intelligent Systems (ELIZA)（Konrad Zuse 学校（ELIZA）） ； Honda Research Institute Europe GmbH, Offenbach, Germany（本田欧洲研究院，奥芬巴赫，德国） ； Uncertainty in Artificial Intelligence Group, TU Eindhoven（人工智能不确定性小组，埃因霍温技术大学） ； Hessian Center for AI (hessian.AI)（黑森人工智能中心（hessian.AI）） ； Center for Cognitive Science（认知科学中心） ； German Center for Artificial Intelligence (DFKI)（德国人工智能中心（DFKI）） 纠错

专题命中 视觉推理 ：VLM(abstract,abstract_cn)；visual reasoning(abstract)；分类 cs.CV、cs.AI、cs.LG

AI总结 提出CycliST基准，通过合成视频评估视频语言模型对循环状态转换的文本推理能力，揭示现有模型在检测循环模式、时间理解和定量分析方面的局限。

Comments Published in the Journal of Data-centric Machine Learning Research (DMLR); https://openreview.net/forum?id=l03g53HUL2

Journal ref Journal of Data-centric Machine Learning Research, 2026

Reward Evolution with Graph-of-Thoughts: A Bi-Level Language Model Framework for Reinforcement Learning

基于思维图的奖励进化：一种用于强化学习的双层语言模型框架

Changwei Yao, Xinzi Liu, Chen Li, Marios Savvides

机构 * Carnegie Mellon University（卡内基梅隆大学） ； University of Tokyo（东京大学） 纠错

专题命中 视觉推理 ：VLM(summary_cn,abstract_cn)；visual language model(abstract)

AI总结 本文提出RE-GoT框架，结合LLM与VLM的图思维推理，通过任务分解和视觉反馈迭代优化奖励函数，实验表明在RoboGen和ManiSkill2任务中均优于现有方法。

Journal ref IEEE International Conference on Robotics and Automation (ICRA 2026)

3DCodeBench: Benchmarking Agentic Procedural 3D Modeling Via Code

3DCodeBench：通过代码进行智能体程序化3D建模的基准测试

Yipeng Gao, Lei Shu, Genzhi Ye, Xi Xiong, Ameesh Makadia, Meiqi Guo, Laurent Itti, Jindong Chen

机构 * Google DeepMind（谷歌DeepMind） ； University of Southern California（南加州大学） ； Google Research（谷歌研究） 纠错

专题命中 视觉推理 ：VLM(abstract,abstract_cn)；vision-language model(abstract)；分类 cs.CV、cs.AI、cs.LG

AI总结 提出3DCodeBench基准，评估12种视觉语言模型将文本和图像参考转换为程序化3D建模代码的能力，并构建基于人类偏好的3DCodeArena排名平台。

Comments Project Page: https://www.3dcodebench.com/; 11 pages (main), with appendix

EMBGuard: Constructing Hazard-Aware Guardrails for Safe Planning in Embodied Agents

EMBGuard：为具身智能体安全规划构建危险感知护栏

Dongwook Choi, Taeyoon Kwon, Bogyung Jeong, Minju Kim, Yeonjun Hwang, Hyojun Kim, Byungchul Kim, Young Kyun Jang, Jinyoung Yeo

机构 * Independent Researcher（独立研究者） ； Department of Biomedical Engineering（生物医学工程系） ； the Department of Intelligent Precision Healthcare Convergence, Sungkyunkwan University（智能精准医疗融合系，全州大学） ； Department of Artificial Intelligence, Yonsei University（人工智能系，延世大学） 纠错

专题命中 视觉推理 ：MLLM(summary_cn,abstract)

AI总结 提出首个基于MLLM的具身安全护栏EMBGuard，通过解耦物理风险推理与智能体策略，评估（视觉观察，动作）对来识别危险配置并提供自然语言解释，同时构建训练数据集EMBHazard和基准测试EMBGuardTest，在紧凑模型尺寸下达到与专有MLLM竞争的性能并降低误报率。

Comments Accepted at ICML 2026

IVR-R1: Refining Trajectories through Iterative Visual-Grounded Reasoning in Reinforcement Learning

IVR-R1：通过强化学习中的迭代视觉基础推理优化轨迹

Chenghao Li, Fusheng Hao, Xikai Zhang, Likang Xiao, Yanwei Ren, Fuxiang Wu, Quan Chen, Liu Liu

机构 * Hangzhou International Innovation Institute, Beihang University（北京航空航天大学杭州国际创新研究院） ； School of Artificial Intelligence, Beihang University（北京航空航天大学人工智能学院） ； Kuaishou Technology（快手科技） ； Shenzhen Institute of Advanced Integration Technology, Shenzhen（深圳先进集成技术研究院） 纠错

专题命中 视觉推理 ：visual reasoning(abstract)；grounding(abstract)；multimodal large language model(abstract)；分类 cs.CV、cs.AI、cs.LG

AI总结 提出IVR-R1框架，利用奖励驱动的筛选机制和迭代再推理循环，在强化学习中动态校正多模态推理轨迹，以解决视觉幻觉和逻辑错误问题。