To address these challenges, the first visual prompting-based multimodal large language model (MLLM) named EarthMarker is proposed in the RS domain. EarthMarker is capable of interpreting RS imagery ...