GeoAgent: Learning to Geolocate Everywhere with Reinforced Geographic Characteristics
This work addresses geolocation tasks for applications requiring fine-grained address conclusions, representing an incremental improvement by refining RL-based methods with domain-specific insights.
The paper tackles the problem of geolocation by introducing GeoAgent, a model that uses expert-annotated data and novel rewards to improve reasoning alignment with geographic characteristics, achieving superior performance over existing methods and general VLLMs across multiple grains.
This paper presents GeoAgent, a model capable of reasoning closely with humans and deriving fine-grained address conclusions. Previous RL-based methods have achieved breakthroughs in performance and interpretability but still remain concerns because of their reliance on AI-generated chain-of-thought (CoT) data and training strategies, which conflict with geographic characteristics. To address these issues, we first introduce GeoSeek, a new geolocation dataset comprising CoT data annotated by geographic experts and professional players. We further thoroughly explore the inherent characteristics of geographic tasks and propose a geo-similarity reward and a consistency reward assessed by a consistency agent to assist training. This encourages the model to converge towards correct answers from a geographic perspective while ensuring the integrity and consistency of its reasoning process. Experimental results show that GeoAgent outperforms existing methods and a series of general VLLMs across multiple grains, while generating reasoning that closely aligns with humans.