← All drillsFrom UCI to Today's Hubs
What you'll be able to do- ✓Fetch a classic UCI dataset and see what its metadata does and does not tell you
- ✓Read a dataset card and a model card on the Hugging Face Hub and check the licence before using either
- ✓Pin a Hub revision and compare it with OpenML's version numbers and checksums
- ✓Put a bootstrap confidence interval on a benchmark score
- ✓Compare a model with a baseline that never sees the premise, and read the gap
- ✓Run an n-gram overlap check for test-set contamination, and test the check on a planted leak
Take the dataset habits from the UCI era to today's hubs, and learn to read a benchmark score the way a sceptic does, with its uncertainty, a simple baseline and an overlap check.
⌁ Choosing a dataset or model to teach with or build on means checking its licence and version and reading its score critically; model cards and dataset cards on hubs such as Hugging Face are where that information lives.
Start this internshipCreate an account to unlock the 4 sections, the workbench, and askThili.
BeginSections
1. Why this course: from the datasets you teach with to the hubs and benchmarks of today
🔒 locked2. The Old Way: One Table from UCI
🔒 locked3. Datasets and Models on the Hub
🔒 locked4. Reading a Benchmark
🔒 lockedDig deeper
🔗UCI Machine Learning Repository: Wine dataset page (licence CC BY 4.0, read 2026-10-26)
article📄Datasheets for Datasets (Gebru et al., 2018)
paper📄Model Cards for Model Reporting (Mitchell et al., 2018)
paper📄A large annotated corpus for learning natural language inference (Bowman et al., 2015): the SNLI benchmark
paper📄Annotation Artifacts in Natural Language Inference Data (Gururangan et al., 2018)
paper