Copertina

Why Do MLLMs Struggle with Spatial Understanding? A Systematic Analysis from Data to Architecture

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Zhang, Wanyue, Huang, Yibin, Xu, Yangbin, Huang, JingJing, Zhi, Helu, Ren, Shuo, Xu, Wang, Zhang, Jiajun
Natura:	Preprint
Pubblicazione:	2025
Soggetti:	Computer Vision and Pattern Recognition
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Video2Layout: Recall and Reconstruct Metric-Grounded Cognitive Map for Spatial Reasoning
di: Huang, Yibin, et al.
Pubblicazione: (2025)

World2VLM: Distilling World Model Imagination into VLMs for Dynamic Spatial Reasoning
di: Zhang, Wanyue, et al.
Pubblicazione: (2026)

Scaling Spatial Reasoning in MLLMs through Programmatic Data Synthesis
di: Helu, Zhi, et al.
Pubblicazione: (2025)

Why MLLMs Struggle to Determine Object Orientations
di: Gopinath, Anju, et al.
Pubblicazione: (2026)

MMPerspective: Do MLLMs Understand Perspective? A Comprehensive Benchmark for Perspective Perception, Reasoning, and Robustness
di: Tang, Yolo Y., et al.
Pubblicazione: (2025)

Spatial Preference Rewarding for MLLMs Spatial Understanding
di: Qiu, Han, et al.
Pubblicazione: (2025)

LEMON: How Well Do MLLMs Perform Temporal Multimodal Understanding on Instructional Videos?
di: Yu, Zhuang, et al.
Pubblicazione: (2026)

MM-UAVBench: How Well Do Multimodal Large Language Models See, Think, and Plan in Low-Altitude UAV Scenarios?
di: Dai, Shiqi, et al.
Pubblicazione: (2025)

Context Tokens are Anchors: Understanding the Repetition Curse in dMLLMs from an Information Flow Perspective
di: Zhao, Qiyan, et al.
Pubblicazione: (2026)

Information Coordination as a Bridge: A Neuro-Symbolic Architecture for Reliable Autonomous Driving Scene Understanding
di: Liu, Shuo, et al.
Pubblicazione: (2026)

STI-Bench: Are MLLMs Ready for Precise Spatial-Temporal World Understanding?
di: Li, Yun, et al.
Pubblicazione: (2025)

On the Generalization Capacities of MLLMs for Spatial Intelligence
di: Zhang, Gongjie, et al.
Pubblicazione: (2026)

3D Spatial Understanding in MLLMs: Disambiguation and Evaluation
di: Chang, Chun-Peng, et al.
Pubblicazione: (2024)

Why Do Vision Language Models Struggle To Recognize Human Emotions?
di: Agarwal, Madhav, et al.
Pubblicazione: (2026)

Do MLLMs Understand Pointing? Benchmarking and Enhancing Referential Reasoning in Egocentric Vision
di: Li, Chentao, et al.
Pubblicazione: (2026)

Why Vision Language Models Struggle with Visual Arithmetic? Towards Enhanced Chart and Geometry Understanding
di: Huang, Kung-Hsiang, et al.
Pubblicazione: (2025)

EgoMind: Activating Spatial Cognition through Linguistic Reasoning in MLLMs
di: Chen, Zhenghao, et al.
Pubblicazione: (2026)

Aesthetic Image Captioning with Saliency Enhanced MLLMs
di: Tao, Yilin, et al.
Pubblicazione: (2025)

EgoExoBench: A Benchmark for First- and Third-person View Video Understanding in MLLMs
di: He, Yuping, et al.
Pubblicazione: (2025)

LRR-Bench: Left, Right or Rotate? Vision-Language models Still Struggle With Spatial Understanding Tasks
di: Kong, Fei, et al.
Pubblicazione: (2025)

ODI-Bench: Can MLLMs Understand Immersive Omnidirectional Environments?
di: Yang, Liu, et al.
Pubblicazione: (2025)

Struct2D: A Perception-Guided Framework for Spatial Reasoning in MLLMs
di: Zhu, Fangrui, et al.
Pubblicazione: (2025)

MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe
di: Yu, Tianyu, et al.
Pubblicazione: (2025)

SPARTUN3D: Situated Spatial Understanding of 3D World in Large Language Models
di: Zhang, Yue, et al.
Pubblicazione: (2024)

SpaceR: Reinforcing MLLMs in Video Spatial Reasoning
di: Ouyang, Kun, et al.
Pubblicazione: (2025)

FullAnno: A Data Engine for Enhancing Image Comprehension of MLLMs
di: Hao, Jing, et al.
Pubblicazione: (2024)

RoadBench: Benchmarking MLLMs on Fine-Grained Spatial Understanding and Reasoning under Urban Road Scenarios
di: Zhang, Jun, et al.
Pubblicazione: (2025)

VideoCap-R1: Enhancing MLLMs for Video Captioning via Structured Thinking
di: Meng, Desen, et al.
Pubblicazione: (2025)

PhysToolBench: Benchmarking Physical Tool Understanding for MLLMs
di: Zhang, Zixin, et al.
Pubblicazione: (2025)

AMBER: An LLM-free Multi-dimensional Benchmark for MLLMs Hallucination Evaluation
di: Wang, Junyang, et al.
Pubblicazione: (2023)

Explore How to Inject Beneficial Noise in MLLMs
di: Zhu, Ruishu, et al.
Pubblicazione: (2025)

VidComposition: Can MLLMs Analyze Compositions in Compiled Videos?
di: Tang, Yolo Y., et al.
Pubblicazione: (2024)

Draw-and-Understand: Leveraging Visual Prompts to Enable MLLMs to Comprehend What You Want
di: Lin, Weifeng, et al.
Pubblicazione: (2024)

Memory Helps, but Confabulation Misleads: Understanding Streaming Events in Videos with MLLMs
di: Zhang, Gengyuan, et al.
Pubblicazione: (2025)

3DRS: MLLMs Need 3D-Aware Representation Supervision for Scene Understanding
di: Huang, Xiaohu, et al.
Pubblicazione: (2025)

MPDrive: Improving Spatial Understanding with Marker-Based Prompt Learning for Autonomous Driving
di: Zhang, Zhiyuan, et al.
Pubblicazione: (2025)

Video-MSR: Benchmarking Multi-hop Spatial Reasoning Capabilities of MLLMs
di: Zhu, Rui, et al.
Pubblicazione: (2026)

Uncovering What, Why and How: A Comprehensive Benchmark for Causation Understanding of Video Anomaly
di: Du, Hang, et al.
Pubblicazione: (2024)

E-VAds: An E-commerce Short Videos Understanding Benchmark for MLLMs
di: Liu, Xianjie, et al.
Pubblicazione: (2026)

Heuristic-inspired Reasoning Priors Facilitate Data-Efficient Referring Object Detection
di: Zhang, Xu, et al.
Pubblicazione: (2026)