VLRS-Bench: A Vision-Language Reasoning Benchmark for Remote Sensing
首个专为遥感领域复杂推理设计的视觉语言基准VLRS-Bench,从认知、决策、预测三维度构建2000个高难度问答对,揭示现有MLLM在遥感推理上的显著瓶颈,为多模态AI在遥感应用的发展提供关键方向。
arXiv:2602.07045v2 Announce Type: replace-cross Abstract: Recent advancements in Multimodal Large Language Models (MLLMs) have enabled complex reasoni…