MapTrace: Scalable Data Generation for Route Tracing on Maps

TL;DR

MapTrace introduces a scalable synthetic data pipeline for route tracing on maps, enabling large-scale fine-tuning of multimodal models to improve their spatial reasoning abilities in map navigation tasks.

Contribution

The paper presents a novel synthetic data generation method for pixel-accurate route annotations, facilitating effective fine-tuning of multimodal models for spatial reasoning.

Findings

Fine-tuning with MapTrace data improves success rates by up to 6.4 points.

Synthetic supervision significantly enhances model robustness in route tracing.

The pipeline enables large-scale, pixel-level annotation for spatial tasks.

Abstract

While Multimodal Large Language Models have achieved human-like performance on many visual and textual reasoning tasks, their proficiency in fine-grained spatial understanding, such as route tracing on maps remains limited. Unlike humans, who can quickly learn to parse and navigate maps, current models often fail to respect fundamental path constraints, in part due to the prohibitive cost and difficulty of collecting large-scale, pixel-accurate path annotations. To address this, we introduce a scalable synthetic data generation pipeline that leverages synthetic map images and pixel-level parsing to automatically produce precise annotations for this challenging task. Using this pipeline, we construct a fine-tuning dataset of 23k path samples across 4k maps, enabling models to acquire more human-like spatial capabilities. Using this dataset, we fine-tune both open-source and proprietary MLLMs. Results on MapBench show that finetuning substantially improves robustness, raising success rates by up to 6.4 points, while also reducing path-tracing error (NDTW). These gains highlight that fine-grained spatial reasoning, absent in pretrained models, can be explicitly taught with synthetic supervision.