Research map / Time series / Foundation models for time series
MIAM: Modality Imbalance-Aware Masking for Multimodal Ecological Applications
Robin Zbinden, Wesley Monteith-Finas, Gencer Sumbul, Nina van Tiel, Chiara Vanalli, Devis Tuia
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 |
|---|---|---|---|
| add relative performance coefficient rho_sm (ablation) | MIAM · GeoPlant validation | AUC (%): 84.8 -> 85.4 | Improved |
| add dynamic learning-speed coefficient rho_dm (ablation) | MIAM · GeoPlant validation | AUC (%): 84.9 -> 85.4 | Improved |
| use non-uniform corner weights instead of uniform (ablation) | MIAM · GeoPlant validation | AUC (%): 85.2 -> 85.4 | Improved |
| use non-uniform corner weights instead of uniform (ablation) | MIAM · TaxaBench validation | top-1 accuracy (%): 36.0 -> 37.1 | Improved |
| use MIAM instead of uniform masking | uniform masking · GeoPlant test | AUC (%): 83.2 -> 86.1 | Improved |
| use MIAM instead of OPM | OPM · GeoPlant test | AUC (%): 83.8 -> 86.1 | Improved |
| use MIAM instead of OPM | OPM · TaxaBench test | top-1 accuracy (%): 31.2 -> 38.7 | Improved |
| use MIAM instead of uniform masking | uniform masking · TaxaBench test | top-1 accuracy (%): 37.7 -> 38.7 | Improved |
| use MIAM masking for SSL pre-training | uniform masking · GeoPlant SSL linear probing | AUC (%): 79.3 -> 79.5 | No clear effect |
| use MIAM instead of OPM in SSL pre-training | OPM · GeoPlant SSL linear probing | AUC (%): 75.3 -> 79.5 | Improved |
| use MIAM on two-modality SatBird | Dirichlet · SatBird | top-30 accuracy: 59.6 -> 60.1 | No clear effect |
| use MIAM instead of oracle-free masking: gap to oracle | MIAM · GeoPlant test | AUC (%): 86.1 vs 87.2 | Mixed |
Closest papers by meaning
- Learning to Rebalance Multi-Modal Optimization by Adaptively Masking Subnetworks (TPAMI 2025 · similarity 0.86)
- ProM3E: Probabilistic Masked MultiModal Embedding Model for Ecology (CVPR 2026 · similarity 0.86)
- MASAM: Multimodal Adaptive Sharpness-Aware Minimization for Heterogeneous Data Fusion (ICLR 2026 · similarity 0.85)
- LDS2AE: Local Diffusion Shared-Specific Autoencoder for Multimodal Remote Sensing Image Classification with Arbitrary Missing Modalities (AAAI 2024 · similarity 0.84)
- Galileo: Learning Global & Local Features of Many Remote Sensing Modalities (ICML 2025 · similarity 0.84)
- Asymmetric Reinforcing Against Multi-Modal Representation Bias (AAAI 2025 · similarity 0.83)
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.
