Decomposing Earth Embeddings with Sparse Autoencoders
cs.LG, cs.CV, physics.geo-ph
Submitted: 2026-11-08
Updated: 2026-11-08
Code: https://github.com/EarthyScience/geosae
Terminology
Sources
- AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data
- Cropland Mapping using Geospatial Embeddings
- Partial recovery of meter-scale surface weather
- Utilizing Earth Foundation Models to Enhance the Simulation Performance of Hydrological Models with AlphaEarth Embeddings
- Leveraging Compact Satellite Embeddings and Graph Neural Networks for Large-Scale Poverty Mapping
- From Pixels to Patches: Pooling Strategies for Earth Embeddings
- OrtSAE: Orthogonal Sparse Autoencoders Uncover Atomic Features
- Discovering Interpretable Biological Concepts in Single-cell RNA-seq Foundation Models
- Towards Open-Ended Visual Scientific Discovery with Sparse Autoencoders
- Interpretable and Testable Vision Features via Sparse Autoencoders
- Towards mechanistic understanding in a data-driven weather model: internal activations reveal interpretable physical features
- Sanity Checks for Sparse Autoencoders: Do SAEs Beat Random Baselines?
- What's in an Earth Embedding? An Explainability Analysis of Location Encoders
- Earth embeddings are decomposable records of the land surface
- Democratizing planetary-scale analysis: An ultra-lightweight Earth embedding database for accurate and flexible global land monitoring
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis