Shortcut Learning in Geospatial Foundation Model Embeddings for Cross-Disaster Damage Assessment
Geospatial foundation models can support disaster damage assessment, but high accuracy on familiar disasters may hide shortcut learning: prediction from event or scene characteristics rather than evidence that transfers to a new disaster.
ProofPaper ↗
Key points
- We study embeddings from TerraMind, Prithvi-EO-2.0, DOFA, and DINOv3 on the multi-disaster datasets xView2 and BRIGHT.
- Four sets of tests examine what information supports prediction: event and acquisition context, before/after image inputs, building count, and correct image pairing.
- Shuffling image pairs within the same disaster and damage class causes modest losses.
- Motivated by these findings, we test whether feature transformations learned only from training disasters improve generalization to unseen disasters.
Sources (1)
- [1]Shortcut Learning in Geospatial Foundation Model Embeddings for Cross-Disaster Damage AssessmentarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 03:50 PM
Geospatial foundation models can support disaster damage assessment, but high accuracy on familiar disasters may hide shortcut learning: prediction from event or scene characteristics rather than evidence that transfers to a new disaster.
We study embeddings from TerraMind, Prithvi-EO-2.0, DOFA, and DINOv3 on the multi-disaster datasets xView2 and BRIGHT.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 9, 2026ConClick: Element-Wise Contrastive Learning Improves GUI Grounding
- Oct 9, 2026Work2Skill: Capturing Human GUI Experience for Reusable Agent Skill Induction
- Oct 8, 2026openai/codex rust-v0.162.0: 0.162.0
- Oct 8, 2026CLUE: A Contrastive Learning Framework for Unsupervised Event Type Induction and Annotation
- Oct 8, 2026Understanding Jev, the new model everyone is talking about
- Oct 7, 2026[AINews] Claude Haiku 5.5 — better than GPT-6 Luna at the same pricing