Thinking with Map: Reinforced Parallel Map-Augmented Agent for Geolocalization

2026-01-08

arXiv.org

AI Summary

This work addresses the underutilization of maps—a critical resource—in image geolocation by current large vision-language models. To bridge this gap, the authors propose the first agent-map interaction framework that centers map-based reasoning as a core cognitive capability, endowing models with “map thinking.” The approach employs a two-stage optimization strategy: first, reinforcement learning enhances the agent’s efficiency in map sampling; second, a parallel test-time expansion mechanism explores multiple candidate paths simultaneously. Evaluated on the MAPBench benchmark, the method significantly improves the Acc@500m metric from 8.0% to 22.1%, substantially outperforming both open-source and closed-source models—including Gemini-3-Pro operating in Google Search/Map mode—thereby demonstrating the effectiveness of map-augmented reasoning in real-world geolocation scenarios.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Computer Vision: Large Vision Models
• Search and Optimization: Learning to Search

Application Category

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

Abstract

The image geolocalization task aims to predict the location where an image was taken anywhere on Earth using visual clues. Existing large vision-language model (LVLM) approaches leverage world knowledge, chain-of-thought reasoning, and agentic capabilities, but overlook a common strategy used by humans -- using maps. In this work, we first equip the model \textit{Thinking with Map} ability and formulate it as an agent-in-the-map loop. We develop a two-stage optimization scheme for it, including agentic reinforcement learning (RL) followed by parallel test-time scaling (TTS). The RL strengthens the agentic capability of model to improve sampling efficiency, and the parallel TTS enables the model to explore multiple candidate paths before making the final prediction, which is crucial for geolocalization. To evaluate our method on up-to-date and in-the-wild images, we further present MAPBench, a comprehensive geolocalization training and evaluation benchmark composed entirely of real-world images. Experimental results show that our method outperforms existing open- and closed-source models on most metrics, specifically improving Acc@500m from 8.0\% to 22.1\% compared to \textit{Gemini-3-Pro} with Google Search/Map grounded mode.

Problem

• image geolocalization
• map-based reasoning
• vision-language models
• geospatial reasoning
• location prediction

Innovation

• Thinking with Map
• agent-in-the-map
• agentic reinforcement learning
• parallel test-time scaling
• MAPBench