Learning to Wander: Improving the Global Image Geolocation Ability of LMMs via Actionable Reasoning

TL;DR

This paper introduces WanderBench, a new geolocation benchmark with interactive panoramas, and GeoAoT, a reasoning framework that actively explores environments to improve image geolocation accuracy.

Contribution

It presents WanderBench for embodied geolocation reasoning and GeoAoT, a novel framework coupling reasoning with physical actions for improved localization.

Findings

GeoAoT outperforms existing models in fine-grained localization.

WanderBench enables evaluation of geolocation in interactive, embodied scenarios.

Models show better generalization in dynamic environments.

Abstract

Geolocation, the task of identifying the geographic location of an image, requires abundant world knowledge and complex reasoning abilities. Though advanced large multimodal models (LMMs) have shown superior aforementioned capabilities, their performance on the geolocation task remains unexplored. To this end, we introduce \textbf{WanderBench}, the first open access global geolocation benchmark designed for actionable geolocation reasoning in embodied scenarios. WanderBench contains over 32K panoramas across six continents, organized as navigable graphs that enable physical actions such as rotation and movement, transforming geolocation from static recognition into interactive exploration. Building on this foundation, we propose \textbf{GeoAoT} (Action of Thought), a \underline{Geo}location framework with \underline{A}ction of \underline{T}hough, which couples reasoning with embodied actions. Instead of generating textual reasoning chains, GeoAoT produces actionable plans such as, approaching landmarks or adjusting viewpoints, to actively reduce uncertainty. We further establish an evaluation protocol that jointly measures geolocation accuracy and difficulty-aware geolocation questioning ability. Experiments on 19 large multimodal models show that GeoAoT achieves superior fine-grained localization and stronger generalization in dynamic environments. WanderBench and GeoAoT define a new paradigm for actionable, reasoning driven geolocation in embodied visual understanding.