Research map / Vision-language and multimodal / Multimodal LLMs
Tesselating The Earth
Daniel Cher, Hamza Iqbal, Eric Xing, Brian Wei, Nathan Jacobs
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 |
|---|---|---|---|
| let Voronoi sites migrate instead of fixing them (ablation) | TTE · geospatial suite | classification accuracy / R2: -12.9% classification, -0.125 regression | Improved |
| add global semantic tokens (ablation) | TTE · geospatial suite | classification accuracy / R2: -3.7% classification, -0.038 regression | Improved |
| add reconstruction loss (ablation) | TTE · EcoRegion / Cali-Housing | accuracy / R2: EcoRegion -1.7, Housing -0.033 | Improved |
| add alignment loss (ablation) | TTE · geospatial suite | classification accuracy / R2 | Improved |
| use peaked image attention instead of symmetric soft attention (ablation) | TTE · geospatial suite | classification accuracy / R2 | Improved |
| double sites from 4,096 to 8,192 (ablation) | TTE · geospatial suite | classification accuracy / R2: 4,096 -> 8,192 sites | No clear effect |
| increase token count beyond R = 64 (ablation) | TTE · geospatial suite | classification accuracy / R2 | Hurt |
| initialize sites from SatCLIP clusters instead of Fibonacci lattice (ablation) | TTE · geospatial suite | average classification: 80.0 -> 80.1 | No clear effect |
| add coordinate conditioning (ablation) | TTE · geospatial suite | average classification: 80.0 -> 77.2 | Mixed |
| let sites move off the sphere (3D sites) (ablation) | TTE · geospatial suite | average classification: 80.0 -> 79.0 | Hurt |
| add RANGE semantic retrieval at inference | TTE · geospatial suite | average classification: 80.0 -> 85.2 | Improved |
| add geodesic retrieval on top of semantic retrieval (ablation) | TTE + RANGE · geospatial suite | average classification: 85.2 -> 84.1 | Hurt |
Closest papers by meaning
- SatCLIP: Global, General-Purpose Location Embeddings with Satellite Imagery (AAAI 2025 · similarity 0.88)
- No Location Left Behind: Measuring and Improving the Fairness of Implicit Representations for Earth Data (ICLR 2025 · similarity 0.88)
- Geographic Location Encoding with Spherical Harmonics and Sinusoidal Representation Networks (ICLR 2024 · similarity 0.87)
- LocDiff: Identifying Locations on Earth by Diffusing in the Hilbert Space (NeurIPS 2025 · similarity 0.87)
- GeoCore-9B: Towards Geo-Aware Generative Foundation Models in Earth Observation (NeurIPS 2026 · similarity 0.86)
- Poly2Vec: Polymorphic Fourier-Based Encoding of Geospatial Objects for GeoAI Applications (ICML 2025 · similarity 0.86)
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.
