EPFL-LARA/lean-compiler-training-corpus download history
EPFL-LARA/lean-compiler-training-corpus is a text classification dataset on the Hugging Face Hub. In the last 30 days it was downloaded 232 times (99 in the last 7 days), and 289 times in total. It ranks #80,131 among datasets by monthly downloads.
Training corpus A compiler-labelled Lean 4 training corpus containing 4,836,380 programs with Lean verdicts and annotated feedback. This is the paper's training corpus, distinct from its in-house validation data and public-prover evaluation benchmark. Labels and feedback use Lean v4.1
Models trained on lean-compiler-training-corpus
3 models list it as training data.
- EPFL-LARA/lean-feedback-qwen3-4b 790 downloads in 30 days
- EPFL-LARA/lean-encoder-modernbert-large 19 downloads in 30 days
- EPFL-LARA/lean-one-token-qwen3-8b 10 downloads in 30 days
Open EPFL-LARA/lean-compiler-training-corpus on Hugging Face
Sister project: Paper Pulse, the upvote history of every Hugging Face Daily Paper.