AI RESEARCH
AgroNVILA: Perception-Reasoning Decoupling for Multi-view Agricultural Multimodal Large Language Models
arXiv CS.AI
•
ArXi:2603.14342v1 Announce Type: cross Agricultural multimodal reasoning requires robust spatial understanding across varying scales, from ground-level close-ups to top-down UAV and satellite imagery. Existing Multi-modal Large Language Models (MLLMs) suffer from a significant "terrestrial-centric" bias, causing scale confusion and logic drift during complex agricultural planning. To address this, we