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

arXiv: https://arxiv.org/abs/2509.02359 · HTML: https://mlsanity.com/paper/2509.02359/

## Key facts

- Cited by 30 papers in the map (data through 14 Oct 2026).
- Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning.
- Builds on 24 of its 38 references that are in the map.
- First posted on arXiv 2 Sep 2025.

## Placed in

- LLM reasoning / [Spatial and embodied reasoning](https://mlsanity.com/lr/spatial-and-embodied-reasoning/)
- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)

## Builds on (most cited first, 24 of 24)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [LLaVA-OneVision: Easy Visual Task Transfer](https://mlsanity.com/paper/2408.03326/) (2024)
- [Visual Genome: Connecting Language and Vision Using Crowdsourced Dense Image Annotations](https://mlsanity.com/paper/1602.07332/) (2016)
- [AI2-THOR: An Interactive 3D Environment for Visual AI](https://mlsanity.com/paper/1712.05474/) (2017)
- [Thinking in Space: How Multimodal Large Language Models See, Remember, and Recall Spaces](https://mlsanity.com/paper/2412.14171/) (2024)
- [What's "up" with vision-language models? Investigating their struggle with spatial reasoning](https://mlsanity.com/paper/2310.19785/) (2023)
- [Rotary Position Embedding for Vision Transformer](https://mlsanity.com/paper/2403.13298/) (2024)
- [Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence](https://mlsanity.com/paper/2505.23747/) (2025)
- [Imagine while Reasoning in Space: Multimodal Visualization-of-Thought](https://mlsanity.com/paper/2501.07542/) (2025)
- [MLLMs Know Where to Look: Training-free Perception of Small Visual Details with Multimodal LLMs](https://mlsanity.com/paper/2502.17422/) (2025)
- [ManipLLM: Embodied Multimodal Large Language Model for Object-Centric Robotic Manipulation](https://mlsanity.com/paper/2312.16217/) (2023)
- [SpaceR: Reinforcing MLLMs in Video Spatial Reasoning](https://mlsanity.com/paper/2504.01805/) (2025)
- [Video-of-Thought: Step-by-Step Video Reasoning from Perception to Cognition](https://mlsanity.com/paper/2501.03230/) (2024)
- [Why Is Spatial Reasoning Hard for VLMs? An Attention Mechanism Perspective on Focus Areas](https://mlsanity.com/paper/2503.01773/) (2025)
- [Mono-InternVL: Pushing the Boundaries of Monolithic Multimodal Large Language Models with Endogenous Visual Pre-training](https://mlsanity.com/paper/2410.08202/) (2024)
- [SpatialCoT: Advancing Spatial Reasoning through Coordinate Alignment and Chain-of-Thought for Embodied Task Planning](https://mlsanity.com/paper/2501.10074/) (2025)
- [Evo-0: Vision-Language-Action Model with Implicit Spatial Understanding](https://mlsanity.com/paper/2507.00416/) (2025)
- [AIC MLLM: Autonomous Interactive Correction MLLM for Robust Robotic Manipulation](https://mlsanity.com/paper/2406.11548/) (2024)
- ["Well, Keep Thinking": Enhancing LLM Reasoning with Adaptive Injection Decoding](https://mlsanity.com/paper/2503.10167/) (2025)
- [Assessing the Value of Visual Input: A Benchmark of Multimodal Large Language Models for Robotic Path Planning](https://mlsanity.com/paper/2507.12391/) (2025)

## Built on by (most cited first, 30 of 30)

- [Spatial-SSRL: Enhancing Spatial Understanding via Self-Supervised Reinforcement Learning](https://mlsanity.com/paper/2510.27606/) (2025)
- [Don't Blind Your VLA: Aligning Visual Representations for OOD Generalization](https://mlsanity.com/paper/2510.25616/) (2025)
- [Multimodal Spatial Reasoning in the Large Model Era: A Survey and Benchmarks](https://mlsanity.com/paper/2510.25760/) (2025)
- [Spatial Reasoning in Multimodal Large Language Models: A Survey of Tasks, Benchmarks and Methods](https://mlsanity.com/paper/2511.15722/) (2025)
- [SpaceVista: All-Scale Visual Spatial Reasoning from mm to km](https://mlsanity.com/paper/2510.09606/) (2025)
- [Look Again, Think Slowly: Enhancing Visual Reflection in Vision-Language Models](https://mlsanity.com/paper/2509.12132/) (2025)
- [SpinBench: Perspective and Rotation as a Lens on Spatial Reasoning in VLMs](https://mlsanity.com/paper/2509.25390/) (2025)
- [Towards Cross-View Point Correspondence in Vision-Language Models](https://mlsanity.com/paper/2512.04686/) (2025)
- [Learning GUI Grounding with Spatial Reasoning from Visual Feedback](https://mlsanity.com/paper/2509.21552/) (2025)
- [The Spatial Blindspot of Vision-Language Models](https://mlsanity.com/paper/2601.09954/) (2026)
- [Self in Space: Benchmarking Self-Awareness and Spatial Cognition in UAV Embodied Intelligence](https://mlsanity.com/paper/2607.12477/) (2026)
- [Video2Layout: Recall and Reconstruct Metric-Grounded Cognitive Map for Spatial Reasoning](https://mlsanity.com/paper/2511.16160/) (2025)
- [Attention in Space: Functional Roles of VLM Heads for Spatial Reasoning](https://mlsanity.com/paper/2603.20662/) (2026)
- [World2VLM: Distilling World Model Imagination into VLMs for Dynamic Spatial Reasoning](https://mlsanity.com/paper/2604.26934/) (2026)
- [Scaling Spatial Reasoning in MLLMs through Programmatic Data Synthesis](https://mlsanity.com/paper/2512.16237/) (2025)
- [Probing Visual Concepts in Lightweight Vision-Language Models for Automated Driving](https://mlsanity.com/paper/2603.06054/) (2026)
- [Why Far Looks Up: Probing Spatial Representation in Vision-Language Models](https://mlsanity.com/paper/2605.30161/) (2026)
- [Loc3R-VLM: Language-based Localization and 3D Reasoning with Vision-Language Models](https://mlsanity.com/paper/2603.18002/) (2026)
- [SSMNBench: Diagnosing Image-based Cross-View Human-Object Understanding via Single-View Sufficiency and Multi-View Necessity](https://mlsanity.com/paper/2606.25634/) (2026)
- [Vision Language Models Are Not (Yet) Spelling Correctors](https://mlsanity.com/paper/2509.17418/) (2025)
- [DynaSolidGeo: A Dynamic Benchmark for Genuine Spatial Mathematical Reasoning of VLMs in Solid Geometry](https://mlsanity.com/paper/2510.22340/) (2025)
- [Predicting Camera Pose from Perspective Descriptions for Spatial Reasoning](https://mlsanity.com/paper/2602.06041/) (2026)
- [Spatial Reasoning is Not a Free Lunch: A Controlled Study on LLaVA](https://mlsanity.com/paper/2603.12545/) (2026)
- [MPDocBench-Parse: Benchmarking Practical Multi-page Document Parsing](https://mlsanity.com/paper/2605.22100/) (2026)
- [SCP: Spatial Causal Prediction in Video](https://mlsanity.com/paper/2603.03944/) (2026)
- [On the Strengths and Weaknesses of Data for Open-set Embodied Assistance](https://mlsanity.com/paper/2603.04819/) (2026)
- [Why MLLMs Struggle to Determine Object Orientations](https://mlsanity.com/paper/2604.13321/) (2026)
- [SVoT: State-aware Visualization-of-Thought for Spatial Reasoning via Reinforcement Learning](https://mlsanity.com/paper/2606.11770/) (2026)
- [ReViCo: Unveiling the Limitations of VLMs in Visual Text Understanding via Error Correction](https://mlsanity.com/paper/2608.27154/) (2026)
- [SpatialCORE: Confidence-Aware Grounded Spatial Reasoning in Large Vision--Language Models](https://mlsanity.com/paper/2609.38716/) (2026)

## Abstract (arXiv)

Spatial understanding is essential for Multimodal Large Language Models (MLLMs) to support perception, reasoning, and planning in embodied environments. Despite recent progress, existing studies reveal that MLLMs still struggle with spatial understanding. However, existing research lacks a comprehensive and systematic evaluation of these limitations, often restricted to isolated scenarios, such as single-view or video. In this work, we present a systematic analysis of spatial understanding from both data and architectural perspectives across three representative scenarios: single-view, multi-view, and video. We propose a benchmark named MulSeT (Multi-view Spatial Understanding Tasks), and design a series of experiments to analyze the spatial reasoning capabilities of MLLMs. From the data perspective, the performance of spatial understanding converges quickly as the training data increases, and the upper bound is relatively low, especially for tasks that require spatial imagination. This indicates that merely expanding training data is insufficient to achieve satisfactory performance. From the architectural perspective, we find that spatial understanding relies more heavily on the positional encoding within the visual encoder than within the language model, in both cascaded and native MLLMs. Moreover, we explore reasoning injection and envision future improvements through architectural design to optimize spatial understanding. These insights shed light on the limitations of current MLLMs and suggest new directions for improving spatial reasoning capabilities through data scaling and architectural tuning.