RSICCLLM: A Multimodal Large Language Model for Remote Sensing Image Change Captioning

2026-06-26

AI Summary

This study addresses the challenges in remote sensing image change captioning, which are primarily constrained by limited model capacity, data scarcity, and difficulties in fine-grained change understanding. To overcome these limitations, this work introduces multimodal large language model post-training to the task for the first time, proposing a difference-aware supervised fine-tuning strategy and a dual-negative-sample preference optimization approach. The authors also construct RSICI—the first instruction-tuning dataset tailored for remote sensing change captioning—and RSICP, a corresponding preference dataset. Remarkably, with only 7 billion parameters, the proposed model substantially outperforms larger-scale baselines, achieving state-of-the-art performance on remote sensing change description. Additionally, the study releases the first dedicated benchmark for evaluating models in this domain.

Abstract

Remote Sensing Image Change Captioning (RSICC) aims to describe changes between bi-temporal remote sensing images and holds significant research and application value. However, most existing methods rely on conventional deep learning architectures, and the limited model capacity constrains performance. Although large-model post-training techniques have achieved great success in general domains, their direct transfer to RSICC remains challenging due to data scarcity and the need for fine-grained change understanding. To address this, we propose RSICCLLM, the first post-training framework for large vision-language models in RSICC. Specifically, we design a data generation paradigm, release the instruction dataset RSICI, and establish a task-specific RSICC benchmark. We further introduce Difference-aware Supervised Fine-tuning to explicitly extract change representations and guide the model in perceiving and understanding temporal differences. In addition, we propose Dual-Negative Preference Optimization (DNPO), which employs two complementary negative-sample construction strategies to construct the preference dataset RSICP and further refine model performance. Extensive experiments validate the superior capability of RSICCLLM, which achieves outstanding results with only 7B parameters, surpassing models of substantially larger scales. The code and dataset will be made publicly available at https://github.com/keaill/RSICCLLM.

Problem

• Remote Sensing Image Change Captioning
• multimodal large language model
• data scarcity
• fine-grained change understanding
• vision-language model

Innovation

• Multimodal Large Language Model
• Remote Sensing Image Change Captioning
• Difference-aware Supervised Fine-tuning
• Dual-Negative Preference Optimization
• Instruction Tuning