Learning to Draw ASCII Improves Spatial Reasoning in Language Models

2026-04-16

AI Summary

Large language models struggle to construct accurate spatial representations from text, limiting their performance in spatial reasoning tasks. This work proposes a method that trains models to translate natural language descriptions into ASCII grid layouts, jointly optimizing layout generation and spatial question answering within token space. The approach uncovers a "read-write asymmetry" phenomenon in language models and demonstrates that explicit layout construction substantially enhances spatial understanding. Experiments show that the model significantly improves the quality of text-to-layout generation on the Text2Space dataset, achieves notably higher accuracy in spatial reasoning, and exhibits strong generalization across three external benchmarks.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Natural Language Processing: (Large) Language Models
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

When faced with complex spatial problems, humans naturally sketch layouts to organize their thinking, and the act of drawing further sharpens their understanding. In this work, we ask whether a similar principle holds for Large Language Models (LLMs): can learning to construct explicit visual layouts from spatial descriptions instill genuine spatial understanding? We introduce Text2Space, a dataset that pairs natural language descriptions with ground-truth ASCII grid layouts and spatial QA pairs, enabling us to separate failures in constructing spatial representations from failures in reasoning over them. We adopt ASCII because it is human-readable, operates entirely within the token space of language models, and encodes spatial relations in a structurally verifiable form. Our evaluation reveals a pronounced "Read-Write Asymmetry": LLMs interpret ASCII representations effectively but struggle to produce them from text, and these construction errors propagate to incorrect answers downstream. To address this limitation, we train models on layout construction (Text$\rightarrow$ASCII) and find that it significantly improves spatial reasoning from text alone, even without producing any ASCII at inference time. Combining construction with comprehension training further amplifies these gains. Crucially, these improvements transfer to three external spatial reasoning benchmarks, demonstrating that, much as sketching sharpens human spatial thinking, learning to construct explicit layouts instills spatial understanding that generalizes beyond the training format.

Problem

• spatial reasoning
• Large Language Models
• spatial representation
• ASCII layouts
• Read-Write Asymmetry

Innovation

• spatial reasoning
• ASCII layout construction
• read-write asymmetry
• Text2Space dataset
• representation learning