RS-WorldModel: a Unified Model for Remote Sensing Understanding and Future Sense Forecasting

2026-03-16

AI Summary

This work addresses the common disconnect in existing remote sensing approaches that treat change understanding and future prediction as isolated tasks, hindering cross-task knowledge transfer. To bridge this gap, the authors propose the first unified world model that jointly models spatiotemporal change comprehension and text-guided future scene generation in remote sensing. They also introduce RSWBench-1.1M, a large-scale language-annotated dataset comprising 1.1 million samples. Through a three-stage training paradigm—Geography-Aware Generative Pretraining (GAGP), Synergistic Instruction Tuning (SIT), and Verifiable Reinforcement Optimization (VRO)—the model achieves state-of-the-art performance with only 2 billion parameters, outperforming open-source models up to 120 times larger on most spatiotemporal change question-answering benchmarks. Moreover, it attains a Fréchet Inception Distance (FID) of 43.13 in future prediction, surpassing all existing open-source models and even closed-source counterparts such as Gemini-2.5-Flash.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Machine Learning: Large Multimodal Models (LMMs)
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Remote sensing world models aim to both explain observed changes and forecast plausible futures, two tasks that share spatiotemporal priors. Existing methods, however, typically address them separately, limiting cross-task transfer. We present RS-WorldModel, a unified world model for remote sensing that jointly handles spatiotemporal change understanding and text-guided future scene forecasting, and we build RSWBench-1.1M, a 1.1 million sample dataset with rich language annotations covering both tasks. RS-WorldModel is trained in three stages: (1) Geo-Aware Generative Pre-training (GAGP) conditions forecasting on geographic and acquisition metadata; (2) synergistic instruction tuning (SIT) jointly trains understanding and forecasting; (3) verifiable reinforcement optimization (VRO) refines outputs with verifiable, task-specific rewards. With only 2B parameters, RS-WorldModel surpasses open-source models up to 120$ \times $ larger on most spatiotemporal change question-answering metrics. It achieves an FID of 43.13 on text-guided future scene forecasting, outperforming all open-source baselines as well as the closed-source Gemini-2.5-Flash Image (Nano Banana).

Problem

• remote sensing
• spatiotemporal change understanding
• future scene forecasting
• world model
• cross-task transfer

Innovation

• unified world model
• spatiotemporal change understanding
• text-guided future forecasting
• synergistic instruction tuning
• verifiable reinforcement optimization