🛰️ Remote Sensing

💬 ACL2026 · 1 paper notes

MONETA: Multimodal Industry Classification through Geographic Information with Multi Agent Systems

The paper proposes MONETA, the first multimodal industry classification benchmark combining text (websites, Wikipedia, Wikidata) and geospatial data (OpenStreetMap, satellite imagery), with zero-shot and multi-turn multi-agent training-free pipelines using open-source and proprietary MLLMs achieving 62.10%-74.10% accuracy on 20-class NACE industry classification, with multi-turn design improving up to 22.80%.