Research questionHow can compact multimodal Earth-observation models handle missing sensors and changing spatial resolutions?Earth-observation systems often receive incomplete combinations of heterogeneous sensor data, while downstream imagery may use resolutions different from pretraining. The challenge is to preserve useful cross-sensor representations without relying on a large parameter budget. Latest papersRecent research connected to this question, newest first.MEOX: Compact Multimodal Mixture-of-Experts for Earth ObservationThe source presents MEOX, a compact multimodal masked autoencoder using sensor-specific adapters, validity signals, sparse shared experts, patch-wise fusion, metadata tokens, and rotary attention. It uses structured sensor dropout during pretraining and reports frozen-transfer results on GEO-Bench at 64- and 224-pixel inputs, along with BigEarthNet finetuning and additional routing, metadata, and retrieval analyses.research paper · Sep 4, 2026