AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

UrbanFusion: Stochastic Multimodal Fusion for Contrastive Learning of Robust Spatial Representations

Dominik J. Mühlematter, Lin Che, Ye Hong, Martin Raubal, Nina Wiedemann

ICML 2026 · Vision-language and multimodal › Cross-modal learning · approach: Multi-modality image fusion · 1 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
train all models on the same PP2-M dataset, use UrbanFusion instead of GAIR or GeoCLIPcoordinate-only location encoding · 8 downstream datasets, linear probeR2 / F1: best on 5 out of 8 datasetsImproved
use UrbanFusion instead of GPS2Veccoordinate-only encoding · Housing Prices and Energy ConsumptionR2Hurt
add multimodal inputs with fused embedding instead of concatenating encodersUrbanFusion multimodal encoding · 6 downstream tasks, linear probeR2 / F1: best on 4 of 6 tasks; lower on land cover and coarse land useMixed
concatenate encoder outputs instead of fusing them (ablation)UrbanFusion · Land cover and coarse land useF1Improved
use multimodal non-coordinate inputs on unseen citiesUrbanFusion · Cross-regional generalization, 6 tasksR2 / F1: ranks first on 5 out of 6 tasksImproved
use MLP downstream learner instead of linear probe (ablation)ZIP code-level tasks · ZIP Code tasksR2Hurt
add latent modality reconstruction loss alongside contrastive loss with random maskingUrbanFusion · synthetic PID datacapture of redundant, unique and synergistic informationImproved
train with only coordinates plus one modality per location (Bimodal) (ablation)UrbanFusion pretraining · all evaluated domainsrelative performance: retains 99.35% of full-pair performanceNo clear effect
train with partially missing modalities (Partial) (ablation)UrbanFusion pretraining · ZIP Code multimodalR2 (%): 75.1 -> 72.3Hurt

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