AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Tesselating The Earth

Daniel Cher, Hamza Iqbal, Eric Xing, Brian Wei, Nathan Jacobs

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Spatial reasoning in VLMs

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 changedWhereMeasuredEffect
let Voronoi sites migrate instead of fixing them (ablation)TTE · geospatial suiteclassification accuracy / R2: -12.9% classification, -0.125 regressionImproved
add global semantic tokens (ablation)TTE · geospatial suiteclassification accuracy / R2: -3.7% classification, -0.038 regressionImproved
add reconstruction loss (ablation)TTE · EcoRegion / Cali-Housingaccuracy / R2: EcoRegion -1.7, Housing -0.033Improved
add alignment loss (ablation)TTE · geospatial suiteclassification accuracy / R2Improved
use peaked image attention instead of symmetric soft attention (ablation)TTE · geospatial suiteclassification accuracy / R2Improved
double sites from 4,096 to 8,192 (ablation)TTE · geospatial suiteclassification accuracy / R2: 4,096 -> 8,192 sitesNo clear effect
increase token count beyond R = 64 (ablation)TTE · geospatial suiteclassification accuracy / R2Hurt
initialize sites from SatCLIP clusters instead of Fibonacci lattice (ablation)TTE · geospatial suiteaverage classification: 80.0 -> 80.1No clear effect
add coordinate conditioning (ablation)TTE · geospatial suiteaverage classification: 80.0 -> 77.2Mixed
let sites move off the sphere (3D sites) (ablation)TTE · geospatial suiteaverage classification: 80.0 -> 79.0Hurt
add RANGE semantic retrieval at inferenceTTE · geospatial suiteaverage classification: 80.0 -> 85.2Improved
add geodesic retrieval on top of semantic retrieval (ablation)TTE + RANGE · geospatial suiteaverage classification: 85.2 -> 84.1Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy