Datasets:
metadata
license: apache-2.0
task_categories:
- text-generation
- feature-extraction
language:
- en
tags:
- theorem-proving
- formal-methods
- lean4
- standard-library
size_categories:
- 10K<n<100K
Lean4-Stdlib
Structured dataset of definitions and theorems from the Lean 4 standard library (Init + Std).
Dataset Description
- Source: leanprover/lean4 (src/Init + src/Std)
- Entries: 40,600
- Files processed: 900 (507 Init + 393 Std)
- License: Apache 2.0
Schema
| Column | Type | Description |
|---|---|---|
| fact | string | Declaration body (without type keyword) |
| type | string | theorem, def, class, structure, instance, inductive, abbrev, axiom, opaque |
| library | string | Init.* or Std.* module |
| imports | list | Import statements |
| filename | string | Source file path |
| symbolic_name | string | Declaration identifier |
| docstring | string | Documentation comment (16% coverage) |
Statistics
By Type
| Type | Count |
|---|---|
| theorem | 31,887 |
| def | 6,600 |
| instance | 426 |
| abbrev | 425 |
| opaque | 399 |
| structure | 382 |
| class | 302 |
| inductive | 163 |
| axiom | 16 |
Docstring Coverage
- 6,651 entries with docstrings (16%)
Use Cases
- Retrieval/RAG for Lean 4
- Learning core Lean 4 patterns
- Training embeddings for formal proofs
- API exploration