GeoAgent: Learning to Geolocate Everywhere with Reinforced Geographic Characteristics

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Jin, Modi, Zhang, Yiming, Sun, Boyuan, Zhang, Dingwen, Cheng, MingMing, Hou, Qibin
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866915796487241728
author Jin, Modi
Zhang, Yiming
Sun, Boyuan
Zhang, Dingwen
Cheng, MingMing
Hou, Qibin
author_facet Jin, Modi
Zhang, Yiming
Sun, Boyuan
Zhang, Dingwen
Cheng, MingMing
Hou, Qibin
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2602_12617
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle GeoAgent: Learning to Geolocate Everywhere with Reinforced Geographic Characteristics
Jin, Modi
Zhang, Yiming
Sun, Boyuan
Zhang, Dingwen
Cheng, MingMing
Hou, Qibin
Artificial Intelligence
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.
title GeoAgent: Learning to Geolocate Everywhere with Reinforced Geographic Characteristics
topic Artificial Intelligence
url https://arxiv.org/abs/2602.12617