Scientific data
Choose data that fits the question and inspect its provenance.
Questions to work through.
Audit benchmark labels with explicit rules
Do compact balance-scale and tic-tac-toe labels agree with their declared arithmetic and board rules?
Open the brief RESEARCH BRIEF · 6 SOURCESBuild a baseline whose preprocessing can be audited
Does a more complicated classifier improve on a simple baseline under the same frozen split and training-only normalization?
Open the briefExplore the sources.
All 176 resources →Iris
Flower measurements for a compact, interpretable multiclass baseline and leakage audit.
2D elastodynamic metamaterials
Pixelated metamaterial unit-cell designs and band-gap locations and widths for investigating simulation surrogate reliability.
3W dataset
Oil-production process signals for studying event detection across operating scenarios.
Abalone
Marine morphometrics for studying relationships between body measurements and growth indicators.
AI4I 2020 Predictive Maintenance Dataset
A documented artificial maintenance dataset for testing rare-failure evaluation and shortcut learning.
Air Quality
Air-quality sensor readings and reference measurements for investigating calibration over time.
Airfoil Self-Noise
Aeroacoustic measurements for examining nonlinear physical response prediction.
Ajwa or Medjool
Date-fruit measurements for comparing interpretable agricultural variety classifiers.
Alcohol QCM Sensor
Quartz-crystal sensor responses for studying chemical discrimination with small repeated-measurement data.
Algerian Forest Fires
Regional weather and fire observations for examining geographic generalization in event prediction.
Amphibians
Habitat descriptors and amphibian occurrence labels for inspecting ecological prediction.
Annealing
Metallurgical process attributes for examining material-category prediction with incomplete fields.