Research map / Vision-language and multimodal / Cross-modal learning
Moonstone: A Multimodal Foundation Model and Benchmark for Lunar Remote Sensing
Ayush Prasad, Swarnalee Mazumder
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use grouped tokenizers instead of per-channel tokenizers (ablation) | MG-MAE · Moonstone geology (linear probe) | accuracy: 49.1 -> 52.4 | Improved |
| add missing-modality key-dimension masking (ablation) | MG-MAE · Moonstone geology (linear probe) | accuracy: 47.8 -> 52.4 | Improved |
| add complementary masking (ablation) | MG-MAE · Moonstone geology (linear probe) | accuracy: 51.0 -> 52.4 | Improved |
| add cross-modal decoder attention (ablation) | MG-MAE · Moonstone geology (linear probe) | accuracy: 50.6 -> 52.4 | Improved |
| add contrastive loss (ablation) | MG-MAE · Moonstone geology (linear probe) | accuracy: 51.3 -> 52.4 | Improved |
| add coverage-adaptive masking (ablation) | MG-MAE · Moonstone crater segmentation | mIoU: 0.685 -> 0.698 | Improved |
| add spectral continuity regularization (ablation) | MG-MAE · Moonstone composition | R2: 0.911 -> 0.924 | Improved |
| use 75% instead of 50% mask ratio (ablation) | MG-MAE · Moonstone geology (linear probe) | accuracy: 51.2 -> 52.4 | Improved |
| use ViT-Base instead of ViT-Small (ablation) | MG-MAE · Moonstone geology (linear probe) | accuracy: 46.5 -> 52.4 | Improved |
| pretrain EO foundation model (Prithvi) instead of scratch | ViT-Base · Moonstone geology (finetuned) | accuracy: 40.1 -> 42.0 | No clear effect |
| pretrain on lunar data with MG-MAE instead of scratch | ViT-Base · Moonstone geology | accuracy: 40.1 -> 56.3 | Improved |
| finetune instead of linear probe | MG-MAE · Moonstone composition | R2: 0.924 -> 0.908 | Hurt |
Closest papers by meaning
- Geometric Foundation Model Distillation for Efficient Lunar 3D Reconstruction (ECCV 2026 · similarity 0.87)
- Mars-Bench: A Benchmark for Evaluating Foundation Models for Mars Science Tasks (NeurIPS 2025 · similarity 0.86)
- Galileo: Learning Global & Local Features of Many Remote Sensing Modalities (ICML 2025 · similarity 0.86)
- GeoLLaVA-8K: Scaling Remote-Sensing Multimodal Large Language Models to 8K Resolution (NeurIPS 2025 · similarity 0.82)
- RoMA: Scaling up Mamba-based Foundation Models for Remote Sensing (NeurIPS 2025 · similarity 0.82)
- REOBench: Benchmarking Robustness of Earth Observation Foundation Models (NeurIPS 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
