# Why Is Spatial Reasoning Hard for VLMs? An Attention Mechanism Perspective on Focus Areas

arXiv: https://arxiv.org/abs/2503.01773 · HTML: https://mlsanity.com/paper/2503.01773/

## Key facts

- Cited by 95 papers in the map (data through 14 Oct 2026).
- Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning.
- Builds on 23 of its 29 references that are in the map.
- First posted on arXiv 3 Mar 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, 23 of 23)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [You Only Look Once: Unified, Real-Time Object Detection](https://mlsanity.com/paper/1506.02640/) (2015)
- [Lost in the Middle: How Language Models Use Long Contexts](https://mlsanity.com/paper/2307.03172/) (2023)
- [Visual Genome: Connecting Language and Vision Using Crowdsourced Dense Image Annotations](https://mlsanity.com/paper/1602.07332/) (2016)
- [VQA: Visual Question Answering](https://mlsanity.com/paper/1505.00468/) (2015)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Language Models (Mostly) Know What They Know](https://mlsanity.com/paper/2207.05221/) (2022)
- [Evaluating Object Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2305.10355/) (2023)
- [Towards VQA Models That Can Read](https://mlsanity.com/paper/1904.08920/) (2019)
- [Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs](https://mlsanity.com/paper/2306.13063/) (2023)
- [Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs](https://mlsanity.com/paper/2406.16860/) (2024)
- [Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs](https://mlsanity.com/paper/2401.06209/) (2024)
- [SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilities](https://mlsanity.com/paper/2401.12168/) (2024)
- [Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding](https://mlsanity.com/paper/2311.16922/) (2023)
- [The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"](https://mlsanity.com/paper/2309.12288/) (2023)
- [Visual Spatial Reasoning](https://mlsanity.com/paper/2205.00363/) (2022)
- [What's "up" with vision-language models? Investigating their struggle with spatial reasoning](https://mlsanity.com/paper/2310.19785/) (2023)
- [G-LLaVA: Solving Geometric Problem with Multi-Modal Large Language Model](https://mlsanity.com/paper/2312.11370/) (2023)
- [Tell Your Model Where to Attend: Post-hoc Attention Steering for LLMs](https://mlsanity.com/paper/2311.02262/) (2023)
- [Unveiling and Harnessing Hidden Attention Sinks: Enhancing Large Language Models without Training through Attention Calibration](https://mlsanity.com/paper/2406.15765/) (2024)
- [In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation](https://mlsanity.com/paper/2403.01548/) (2024)

## Built on by (most cited first, 30 of 95)

- [OmniSpatial: Towards Comprehensive Spatial Reasoning Benchmark for Vision Language Models](https://mlsanity.com/paper/2506.03135/) (2025)
- [Seeing but Not Believing: Probing the Disconnect Between Visual Attention and Answer Correctness in VLMs](https://mlsanity.com/paper/2510.17771/) (2025)
- [How Far are VLMs from Visual Spatial Intelligence? A Benchmark-Driven Perspective](https://mlsanity.com/paper/2509.18905/) (2025)
- [Visual Representation Alignment for Multimodal Large Language Models](https://mlsanity.com/paper/2509.07979/) (2025)
- [Beyond Semantics: Rediscovering Spatial Awareness in Vision-Language Models](https://mlsanity.com/paper/2503.17349/) (2025)
- [Multimodal Spatial Reasoning in the Large Model Era: A Survey and Benchmarks](https://mlsanity.com/paper/2510.25760/) (2025)
- [Why Do MLLMs Struggle with Spatial Understanding? A Systematic Analysis from Data to Architecture](https://mlsanity.com/paper/2509.02359/) (2025)
- [DecAlign: Hierarchical Cross-Modal Alignment for Decoupled Multimodal Representation Learning](https://mlsanity.com/paper/2503.11892/) (2025)
- [Spatial Reasoning in Multimodal Large Language Models: A Survey of Tasks, Benchmarks and Methods](https://mlsanity.com/paper/2511.15722/) (2025)
- [Thinking with Blueprints: Assisting Vision-Language Models in Spatial Reasoning via Structured Object Representation](https://mlsanity.com/paper/2601.01984/) (2026)
- [CompoDistill: Attention Distillation for Compositional Reasoning in Multimodal LLMs](https://mlsanity.com/paper/2510.12184/) (2025)
- [Linear Mechanisms for Spatiotemporal Reasoning in Vision Language Models](https://mlsanity.com/paper/2601.12626/) (2026)
- [TwinBrainVLA: Unleashing the Potential of Generalist VLMs for Embodied Tasks via Asymmetric Mixture-of-Transformers](https://mlsanity.com/paper/2601.14133/) (2026)
- [PhysicsMind: Sim and Real Mechanics Benchmarking for Physical Reasoning and Prediction in Foundational VLMs and World Models](https://mlsanity.com/paper/2601.16007/) (2026)
- [Tell Model Where to Look: Mitigating Hallucinations in MLLMs by Vision-Guided Attention](https://mlsanity.com/paper/2511.20032/) (2025)
- [V-SEAM: Visual Semantic Editing and Attention Modulating for Causal Interpretability of Vision-Language Models](https://mlsanity.com/paper/2509.14837/) (2025)
- [Mind Meets Space: Rethinking Agentic Spatial Intelligence from a Neuroscience-inspired Perspective](https://mlsanity.com/paper/2509.09154/) (2025)
- [CausalSpatial: A Benchmark for Object-Centric Causal Spatial Reasoning](https://mlsanity.com/paper/2601.13304/) (2026)
- [[De|Re]constructing VLMs' Reasoning in Counting](https://mlsanity.com/paper/2510.19555/) (2025)
- [Learning Modal-Mixed Chain-of-Thought Reasoning with Latent Embeddings](https://mlsanity.com/paper/2602.00574/) (2026)
- [InfiniBench: Infinite Benchmarking for Visual Spatial Reasoning with Customizable Scene Complexity](https://mlsanity.com/paper/2511.18200/) (2025)
- [Vision to Geometry: 3D Spatial Memory for Sequential Embodied MLLM Reasoning and Exploration](https://mlsanity.com/paper/2512.02458/) (2025)
- [Video-MSR: Benchmarking Multi-hop Spatial Reasoning Capabilities of MLLMs](https://mlsanity.com/paper/2601.09430/) (2026)
- [Gaze Heads: How VLMs Look at What They Describe](https://mlsanity.com/paper/2606.14703/) (2026)
- [Pathways of Visual Information Flow in Vision-Language Models](https://mlsanity.com/paper/2607.03358/) (2026)
- [Caption This, Reason That: VLMs Caught in the Middle](https://mlsanity.com/paper/2505.21538/) (2025)
- [SCALE: Self-uncertainty Conditioned Adaptive Looking and Execution for Vision-Language-Action Models](https://mlsanity.com/paper/2602.04208/) (2026)
- [Attention in Space: Functional Roles of VLM Heads for Spatial Reasoning](https://mlsanity.com/paper/2603.20662/) (2026)
- [The Dual Mechanisms of Spatial Variable Binding in Vision-Language Models](https://mlsanity.com/paper/2603.22278/) (2026)
- [World2VLM: Distilling World Model Imagination into VLMs for Dynamic Spatial Reasoning](https://mlsanity.com/paper/2604.26934/) (2026)

## Abstract (arXiv)

Large Vision Language Models (VLMs) have long struggled with spatial reasoning tasks. Surprisingly, even simple spatial reasoning tasks, such as recognizing "under" or "behind" relationships between only two objects, pose significant challenges for current VLMs. In this work, we study the spatial reasoning challenge from the lens of mechanistic interpretability, diving into the model's internal states to examine the interactions between image and text tokens. By tracing attention distribution over the image through out intermediate layers, we observe that successful spatial reasoning correlates strongly with the model's ability to align its attention distribution with actual object locations, particularly differing between familiar and unfamiliar spatial relationships. Motivated by these findings, we propose ADAPTVIS based on inference-time confidence scores to sharpen the attention on highly relevant regions when confident, while smoothing and broadening the attention window to consider a wider context when confidence is lower. This training-free decoding method shows significant improvement (e.g., up to a 50 absolute point improvement) on spatial reasoning benchmarks such as WhatsUp and VSR with negligible cost. We make code and data publicly available for research purposes at https://github.com/shiqichen17/AdaptVis.