AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Measuring the Intrinsic Dimension of Earth Representations

Arjun Rao, Marc Rußwurm, Konstantin Klemmer, Esther Rolf

ICLR 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory · 9 citations (OpenAlex)

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
increase SatCLIP spherical-harmonic order from L=10 to L=40 (ablation)SatCLIP location encoder · Land sampling (100k points)global ID (FisherS): 5.00 -> 8.08Improved
increase SatCLIP embedding size D from 64 to 512 (ablation)SatCLIP-L10 · Embedding-size studyglobal ID (MLE): 1.958 -> 1.956No clear effect
increase SatCLIP embedding size D from 64 to 512 (ablation)SatCLIP-L10 MLP head · Air temperatureR2: 95.9 -> 95.9No clear effect
increase RCF embedding size D from 64 to 512 (ablation)RCF image encoder · S2-100Kglobal ID (MLE): 5.56 -> 6.32No clear effect
increase GeoCLIP max RFF frequency and hierarchy levels M (ablation)GeoCLIP location encoder · YFCC fine-tuningglobal ID (FisherS)Improved
increase number of multi-scale components S (ablation)Space2Vec / Sphere2Vec · MOSAIKS nightlightsglobal IDImproved
pre-train SatCLIP on more modalities (S2 to S1+S2 to all MMEarth) (ablation)SatCLIP · MMEarth air temp, elevation, populationglobal ID and R2Improved
higher ID of frozen pretrained embeddingsgeographic INRs · five TorchSpatial taskstest R2 / top-1Improved
lower TwoNN ID of head activationsMLP head on frozen INR · five TorchSpatial taskstest R2 / top-1Improved
end-to-end supervised training instead of pretrainingSphere2Vec and Space2Vec · SustainBench image-location regressionR2Improved
switch estimator from distance-based MLE to angle-based FisherSSatCLIP-L40 · Land sampling (100k points)global ID: 2.03 -> 8.08Mixed
increase neighbors k from 5 to 200 with ESS (ablation)GeoCLIP · Fibonacci samplingglobal ID (ESS): 12.031 -> 98.897Hurt

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