MEOX is a compact multimodal masked autoencoder designed for Earth Observation that uses a 2.939 million‑parameter encoder and 3.115 million total parameters. It incorporates sensor‑specific adapters, explicit validity signals, and a shared sparse‑expert block to maintain modality‑dependent processing before a learned patch‑wise fusion, followed by fourteen encoder blocks that process a single spatial sequence with four metadata tokens. Pretrained on 1.228 million MMEarth64 samples, MEOX achieves strong performance on GEO‑Bench tasks, surpassing prior CSMoE results, and demonstrates effective sensor‑flexible representation learning with a modest parameter budget.
arXiv:2606. 10819v1 Announce Type: cross Abstract: RS-MLLMs enable natural-language understanding and spatial reasoning over earth observation imagery.
arXiv:2608. 06612v1 Announce Type: cross Abstract: The plethora of readily available geospatial data offers exciting opportunities to learn high quality representations of the planet, but the sheer size of the Earth Observations (EO), differing modalities, and different sensor types pose significant challenges in doing so.
arXiv:2606. 12595v1 Announce Type: cross Abstract: Foundation models are rapidly transforming Earth observation by enabling scalable pretraining across diverse unlabeled geospatial modalities.
The paper introduces a multimodal foundation model for lunar remote sensing, trained from scratch on SomBench—a dataset of nearly two million co‑registered tile bundles across 11 modalities at 1 m and 100 m resolutions. The model extends the TerraMind masked‑token architecture with lunar‑specific features such as explicit acquisition geometry and joint training of two spatial scales, and employs FlexiViT patch embeddings for adaptable patch sizes. Evaluation on crater detection, irregular mare patch segmentation, and polar ice prospectivity regression shows that the pretrained model matches or surpasses ImageNet‑pretrained baselines, with notable label efficiency and effective adaptation via LoRA.
arXiv:2607. 03644v1 Announce Type: cross Abstract: Decades of orbital missions have produced multi-modal remote sensing data for the Moon, spanning optical imagery, spectroscopy, thermal emission, radar, gravity, and elemental composition.
arXiv:2606. 13896v1 Announce Type: cross Abstract: Self-supervised geospatial foundation models (GeoFMs) learn transferable representations from remote sensing data, but their downstream behavior is difficult to characterize.
arXiv:2504. 11171v5 Announce Type: replace-cross Abstract: We present TerraMind, the first any-to-any generative, multimodal foundation model for Earth observation (EO).
SIMPLER is a pre‑fine‑tuning method that reduces inference and deployment costs for Earth Observation foundation models by pruning redundant layers. It uses layer‑wise representation similarity on unlabeled task data to identify and remove up to 79% of parameters without requiring gradients, magnitude heuristics, or hyperparameter tuning. Experiments on Prithvi‑EO‑2, TerraMind, and ImageNet‑pretrained ViT‑MAE show that SIMPLER retains 94% of baseline performance while achieving 2.1× faster training and 2.6× faster inference.
arXiv:2606. 14760v1 Announce Type: cross Abstract: Remote-sensing foundation models (RSFMs) benefit from pretraining on imagery from multiple sensors and ground sampling distances (GSDs), but such exposure alone does not resolve scale mismatch during downstream adaptation.
arXiv:2608. 19766v1 Announce Type: cross Abstract: Self-supervised pretraining on remote sensing imagery typically treats all samples as equally informative, despite large variability in geographic and visual structure.
DCA-MoE is a crowd‑counting framework that keeps a frozen DINOv3 encoder while introducing two content‑dependent modules: Spatially Adaptive Layer Fusion (SALF) assigns position‑wise weights to four backbone features, and Density‑Routed Multi‑Receptive‑Field Experts (DR‑MoE) mixes local, mid‑range, and large‑context residual experts for each location. An EBC‑style head reconstructs block density, and the model is trained with DMCount supervision plus a routing‑balance term. On the NWPU‑Crowd validation split, the best configuration based on DINOv3 ViT‑L/16 achieves 31.7 MAE and 72.2 RMSE, while the ViT‑B/16 full model records 32.2/75.9.
arXiv:2607. 18504v1 Announce Type: cross Abstract: Benchmarks for Geospatial Foundation Models (GFMs) increasingly rank models by aggregate score, but such rankings obscure why models differ: how much of the gap is architecture, how much is decoder capacity, and how much is a use-case-specific artefact?