# From Pixels to BFS: High Maze Accuracy Does Not Imply Visual Planning

arXiv: https://arxiv.org/abs/2603.26839 · HTML: https://mlsanity.com/paper/2603.26839/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 12 of its 19 references that are in the map.
- First posted on arXiv 27 Mar 2026.

## Builds on (most cited first, 12 of 12)

- [MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI](https://mlsanity.com/paper/2311.16502/) (2023)
- [Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters](https://mlsanity.com/paper/2408.03314/) (2024)
- [MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts](https://mlsanity.com/paper/2310.02255/) (2023)
- [MathVerse: Does Your Multi-modal LLM Truly See the Diagrams in Visual Math Problems?](https://mlsanity.com/paper/2403.14624/) (2024)
- [BLINK: Multimodal Large Language Models Can See but Not Perceive](https://mlsanity.com/paper/2404.12390/) (2024)
- [Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey](https://mlsanity.com/paper/2503.12605/) (2025)
- [Can MLLMs Reason in Multimodality? EMMA: An Enhanced MultiModal ReAsoning Benchmark](https://mlsanity.com/paper/2501.05444/) (2025)
- [Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens](https://mlsanity.com/paper/2511.19418/) (2025)
- [SpatialBench: Benchmarking Multimodal Large Language Models for Spatial Cognition](https://mlsanity.com/paper/2511.21471/) (2025)
- [GSR-BENCH: A Benchmark for Grounded Spatial Reasoning Evaluation via Multimodal LLMs](https://mlsanity.com/paper/2406.13246/) (2024)
- [MMEvalPro: Calibrating Multimodal Benchmarks Towards Trustworthy and Efficient Evaluation](https://mlsanity.com/paper/2407.00468/) (2024)
- [SpatiaLab: Can Vision-Language Models Perform Spatial Reasoning in the Wild?](https://mlsanity.com/paper/2602.03916/) (2026)

## Built on by (most cited first, 1 of 1)

- [Spatial Strategies, Not Actions: Vector-Quantized Geodesics as Tools for LLM-Driven Agents](https://mlsanity.com/paper/2610.00613/) (2026)

## Abstract (arXiv)

How do multimodal models solve visual spatial tasks -- through genuine planning, or through brute-force search in token space? We introduce textscMazeBench, a benchmark of 110 procedurally generated maze images across nine controlled groups, and evaluate 16 model configurations from OpenAI, Anthropic, Google, and Alibaba. GPT-5.4 solves 91% and Gemini 3.1 Pro 79%, but these scores are misleading: models typically translate images into text grids and then enumerate paths step by step, consuming 1,710--22,818 tokens per solve for a task humans do quickly. Without added reasoning budgets, all configurations score only 2--12%; on 20×20 ultra-hard mazes, they hit token limits and fail. Qualitative traces reveal a common two-stage strategy: image-to-grid translation followed by token-level search, effectively BFS in prose. A text-grid ablation shows Claude Sonnet 4.6 rising from 6% on images to 80% when given the correct grid, isolating weak visual extraction from downstream search. When explicitly instructed not to construct a grid or perform graph search, models still revert to the same enumeration strategy. textscMazeBench therefore shows that high accuracy on visual planning tasks does not imply human-like spatial understanding.