Research map / Label-efficient and robust learning / Test-time adaptation
WILDS: A Benchmark of in-the-Wild Distribution Shifts
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use CORAL to align feature means and covariances across domains | ERM on iWildCam2020-WILDS · iWildCam2020-WILDS | Macro F1: 31.0 (1.3) -> 32.8 (0.1) | Helped |
| Optimize for the highest-loss training domain with Group DRO | ERM on Camelyon17-WILDS · Camelyon17-WILDS | Average accuracy: 70.3 (6.4) -> 68.4 (7.3) | Hurt |
| Use CORAL to align feature means and covariances across domains | ERM on RxRx1-WILDS · RxRx1-WILDS | Average accuracy: 29.9 (0.4) -> 28.4 (0.3) | Hurt |
| Optimize for the highest-loss training domain with Group DRO | ERM on OGB-MolPCBA · OGB-MolPCBA | Average AP: 27.2 (0.3) -> 22.4 (0.6) | Hurt |
| Optimize for the highest-loss training domain with Group DRO | ERM on GlobalWheat-WILDS · GlobalWheat-WILDS | Average domain accuracy: 49.2 (1.5) -> 46.1 (1.6) | Hurt |
| Optimize for the highest-loss training domain with Group DRO | ERM on CivilComments-WILDS · CivilComments-WILDS | Worst-group accuracy: 56.0 (3.6) -> 70.0 (2.0) | Helped |
| Use CORAL to align feature means and covariances across domains | ERM on FMoW-WILDS · FMoW-WILDS | Worst-region accuracy: 32.3 (1.3) -> 31.7 (1.2) | Hurt |
| Use CORAL to align feature means and covariances across domains | ERM on PovertyMap-WILDS · PovertyMap-WILDS | Worst-U/R Pearson R: 0.45 (0.06) -> 0.44 (0.06) | Hurt |
| Use CORAL to align feature means and covariances across domains | ERM on Amazon-WILDS · Amazon-WILDS | 10th percentile accuracy: 53.8 (0.8) -> 52.9 (0.8) | Hurt |
| Use CORAL to align feature means and covariances across domains | ERM on Py150-WILDS · Py150-WILDS | Method/class accuracy: 67.9 (0.1) -> 65.9 (0.1) | Hurt |
Closest papers by meaning
- "What is Different Between These Datasets?" A Framework for Explaining Data Distribution Shifts (NeurIPS 2026 · similarity 0.85)
- Benchmarking Object Detectors under Real-World Distribution Shifts in Satellite Imagery (CVPR 2025 · similarity 0.85)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 2024 · similarity 0.84)
- Multiply Robust Estimation for Local Distribution Shifts with Multiple Domains (ICML 2024 · similarity 0.84)
- Treasure Hunt: Real-time Targeting of the Long Tail using Training-Time Markers (NeurIPS 2025 · similarity 0.84)
- DivShift: Exploring Domain-Specific Distribution Shifts in Large-Scale, Volunteer-Collected Biodiversity Datasets (AAAI 2025 · similarity 0.84)
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.