SustainBench

Sustainability benchmark suite across agriculture, poverty, land cover, water, climate-action, and related geospatial ML tasks.

18rows
task_scoreprimary metric
2026-05-27sampled

Metadata

Metrics

R2, F1 Pixel, F1 Image, Macro F1, Accuracy, RMSE (lower is better), Dice, Kappa

Latest Results

Rows parsed from SustainBench task-specific leaderboard tables. Each row is a task-level result because the source mixes metrics across sustainability tasks.

Rank Subject Score Model Match Provenance Sampled
1 Baseline: KNN (Poverty prediction over space) 0.63 Imported 2026-05-27
2 Baseline: modified ResNet-18 using all satellite bands (Poverty prediction over time) 0.35 Imported 2026-05-27
1 U-Net (Weakly supervised cropland classification) 0.88 Imported 2026-05-27
1 Rustowicz et al. (Crop type classification) 57.3 Imported 2026-05-27
1 Rustowicz et al. (Crop type classification) 69.7 Imported 2026-05-27
1 You et al. (Crop yield prediction) 0.37 Imported 2026-05-27
1 Wang et al. (Crop yield prediction) 0.62 Imported 2026-05-27
1 Wang et al. (Crop yield prediction) 0.42 Imported 2026-05-27
1 Aung et al. (Field delineation) 0.61 Imported 2026-05-27
1 Baseline: KNN (Child mortality rate) 0.01 Imported 2026-05-27
1 Baseline: KNN (Women BMI) 0.42 Imported 2026-05-27
12 GCN (Lee et al.) (Women BMI) 0.57 (India) Imported 2026-05-27
1 Baseline: KNN (Women educational attainment) 0.26 Imported 2026-05-27
1 Baseline: KNN (Water quality index) 0.40 Imported 2026-05-27
1 Baseline: KNN (Sanitation index) 0.36 Imported 2026-05-27
16 Lee et al. (Brick kiln detection) 94.2% Imported 2026-05-27
1 Tile2Vec with ResNet-50 (Representation learning for land cover) 0.55 (n= 1,000) 0.58 (n= 10,000) Imported 2026-05-27
1 MAML with shallow 1D CNN (Out-of-domain land cover classification) 0.32 (1-shot, 2-way) Imported 2026-05-27