# Google Research — Teaching AI to read a map

- Company: Google Research (research.google)
- Announced: 2026-02-17T21:37:00+00:00
- Category: research-paper
- Subject: Research
- Models affected: Gemini 2.5 Pro, Imagen-4
- Source: https://research.google/blog/teaching-ai-to-read-a-map/
- Record: https://forck.live/items/1365-teaching-ai-to-read-a-map

Google Research introduces MapTrace, a synthetic data generation pipeline and dataset of 2 million question-answer pairs to teach multimodal large language models to trace valid paths on maps, addressing a gap in spatial reasoning. The pipeline uses Gemini 2.5 Pro and Imagen-4 models to generate diverse maps and produce pixel-level annotations.

## Evidence

Verbatim from https://research.google/blog/teaching-ai-to-read-a-map/:

> To address this challenge, in “MapTrace: Scalable Data Generation for Route Tracing on Maps”, we introduce a new task, dataset, and synthetic data generation pipeline designed to teach MLLMs the fundamental skill of tracing paths on maps. ... We also open-source 2M question answer pairs generated with the proposed pipeline utilizing Gemini 2.5 Pro and Imagen-4 Models to encourage the research community to further explore this area.

---

Record: https://forck.live/items/1365-teaching-ai-to-read-a-map
Catalogue: https://forck.live/llms.txt
Feed: https://forck.live/feed.md