Where Do Vision Language Models Fail at 3D Reasoning?
Khai Le-Duc, Nguyen Kim Hai Bui, An Ngo, Nam D. Hoang, Yueqi Song, Jiarui Liu, Thai Quoc Hoang, Fang Wu, Chris Ngo, Terry Jingchen Zhang, Mona T. Diab, Ran Xu, … 展开作者
Khai Le-Duc, Nguyen Kim Hai Bui, An Ngo, Nam D. Hoang, Yueqi Song, Jiarui Liu, Thai Quoc Hoang, Fang Wu, Chris Ngo, Terry Jingchen Zhang, Mona T. Diab, Ran Xu, Silvio Savarese, Xiao-Ming Wu
Vision-language models can get individual judgments right while struggling to recover and consistently use the exact properties of a 3D object. To characterize this gap, we introduce BedBench, a controlled benchmark of thirteen procedurally generated tasks spanning individual judgments and composite answers. Across twelve open-weight and four frontier models, per-color counting exposes a pronounced weakness: humans achieve 93.3% accuracy, compared with 7.2% for the best model. Targeted fine-tuning raises one model's accuracy from 3.0% to 68.6%, but reveals a further difficulty: its answers about the most common colors remain inconsistent with its own per-color counts. Supplying those counts as text improves its judgments, showing that recovering a property does not mean using it consistently. Reasoning, scaling, and added visual cues each improve individual judgments, but none makes exact answers reliable. To locate the difficulty, we probe Qwen3.5-27B: exact per-color counts remain recoverable at the language model's input, yet the model's answers rarely recover them (65.0% for the probe vs. 3.0% for the model). Image-removal controls further show that some apparent successes require no visual evidence. These findings motivate evaluating 3D reasoning by exact correctness, consistent use, and image dependence, rather than accuracy alone. Code and dataset are publicly available: https://github.com/leduckhai/BedBench
Locatability-Guided Adaptive Reasoning for Image Geo-Localization with Vision-Language Models
The University of Utah(犹他大学);The University of Oklahoma(俄克拉荷马大学);Worcester Polytechnic Institute(沃斯特理工学院);Rensselaer Polytechnic Institute(雷士利理工学院);University of Illinois Chicago(伊利诺伊大学芝加哥分校)
MultihopSpatial: Multi-hop Compositional Spatial Reasoning Benchmark for Vision-Language Model
Electronics and Telecommunications Research Institute, South Korea(韩国电信研究院);Korea Advanced Institute of Science and Technology, South Korea(韩国科学技术院);Sungkyunkwan University, South Korea(成均馆大学);DeepAuto, South Korea(DeepAuto)
ViSA-Enhanced Aerial VLN: A Visual-Spatial Reasoning Enhanced Framework for Aerial Vision-Language Navigation
Tianmushan Laboratory, Beihang University(北航之梦实验室,北京航空航天大学);Hangzhou International Innovation Institute, Beihang University(杭州国际创新研究院,北京航空航天大学);Foshan Graduate School of Innovation, Northeastern University(佛山创新研究生学院,东北大学);School of Aeronautic Science and Engineering, Beihang University(航空科学与工程学院,北京航空航天大学);Faculty of Robot Science and Engineering, Northeastern University(机器人科学与工程学院,东北大学)
Reasoning in the Dark: Interleaved Vision-Text Reasoning in Latent Space
The Hong Kong Polytechnic University(香港理工大学);Singapore Management University(新加坡管理学院);Shandong University(山东大学);Harbin Institute of Technology (Shenzhen)(哈尔滨工业大学(深圳))
R3G: A Reasoning-Retrieval-Reranking Framework for Vision-Centric Answer Generation
The Shenzhen International Graduate School, Tsinghua University, China(清华大学深圳国际研究生院);State Key Laboratory of Nuclear Power Safety Technology and Equipment, China(核能安全技术与装备国家重点实验室);School of Computer Science and Information Engineering, Hefei University of Technology, China(合肥工业大学计算机科学与信息工程学院)
ReasoningTrack: Chain-of-Thought Reasoning for Long-term Vision-Language Tracking
School of Computer Science and Technology, Anhui University(安徽大学计算机科学与技术学院);School of Artificial Intelligence, Shanghai Jiao Tong University(上海交通大学人工智能学院);School of Electronic and Information Engineering, Anhui University(安徽大学电子与信息工程学院)
RAVEN: A Dataset for Relational and Analogical Visual rEasoNing
UCLA Center for Vision, Cognition, Learning and Autonomy(加州大学洛杉矶分校视觉、认知、学习与自主性中心);International Center for AI and Robot Autonomy (CARA)(人工智能与机器人自主性国际中心(CARA))
Unveiling Chain of Step Reasoning for Vision-Language Models with Fine-grained Rewards
Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);School of Artificial Intelligence, University of Chinese Academy of Sciences(中国科学院大学人工智能学院);Beijing Academy of Artificial Intelligence(北京人工智能研究院)
Can MLLMs Reason Beyond Language? VisReason: A Comprehensive Benchmark for Vision-Centric Reasoning
Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);School of Artificial Intelligence, University of Chinese Academy of Sciences(中国科学院大学人工智能学院)
Position: Reasoning After Perception Means Reasoning Without Vision
Tsinghua University(清华大学);Peking University(北京大学);National University of Singapore(新加坡国立大学);Zhongguancun Academy(中关村学院);Nanjing University(南京大学);Nankai University(南开大学)
Proof-of-TBI -- Fine-Tuned Vision Language Model Consortium and OpenAI-o3 Reasoning LLM-Based Medical Diagnosis Support System for Mild Traumatic Brain Injury (TBI) Prediction
Old Dominion University(旧 Dominion 大学);McDonald Army Health Center(麦克唐纳陆军医疗中心);BRAINBox Solutions(BRAINBox 解决方案);UVA Health(弗吉尼亚大学健康系统)
VReST: Enhancing Reasoning in Large Vision-Language Models through Tree Search and Self-Reward Mechanism
School of Computer Science and Engineering, Sun Yat-Sen University(中山大学计算机科学与工程学院);Guangdong Laboratory of Artificial Intelligence and Digital Economy(广东省人工智能与数字经济实验室);Huawei Technologies Co., Ltd(华为技术有限公司)
Vlaser: Vision-Language-Action Model with Synergistic Embodied Reasoning
University of Science and Technology of China(中国科学技术大学);Shanghai AI Laboratory(上海人工智能实验室);Shanghai Jiao Tong University(上海交通大学);Zhejiang University(浙江大学);Nanjing University(南京大学);Fudan University(复旦大学);Tsinghua University(清华大学);NUS(新加坡国立大学);Northeastern University(东北大学);Shenzhen University(深圳大学)