metadata
license: mit
task_categories:
- text-generation
- feature-extraction
language:
- en
tags:
- theorem-proving
- formal-methods
- agda
- standard-library
- dependent-types
size_categories:
- 1K<n<10K
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
dataset_info:
features:
- name: fact
dtype: string
- name: type
dtype: string
- name: library
dtype: string
- name: imports
list: string
- name: filename
dtype: string
- name: symbolic_name
dtype: string
- name: docstring
dtype: string
splits:
- name: train
num_bytes: 4138264
num_examples: 7074
download_size: 300725
dataset_size: 4138264
Agda-Stdlib
Structured dataset of definitions and types from the Agda standard library v2.3.
Dataset Description
- Source: agda/agda-stdlib (v2.3)
- Entries: 7,074
- Files processed: 1,153
- License: MIT
Schema
| Column | Type | Description |
|---|---|---|
| fact | string | Type signature and definition |
| type | string | function, data, record |
| library | string | Top-level module (Data, Relation, Algebra, etc.) |
| imports | list | Import statements |
| filename | string | Source file path |
| symbolic_name | string | Declaration identifier |
| docstring | string | Documentation comment (3% coverage) |
Statistics
By Type
| Type | Count |
|---|---|
| function | 6,804 |
| record | 208 |
| data | 62 |
By Library (Top 10)
| Library | Count |
|---|---|
| Data | 4,100 |
| Relation | 791 |
| Algebra | 741 |
| Function | 433 |
| Codata | 358 |
| Reflection | 178 |
| Effect | 113 |
| System | 92 |
| Text | 86 |
| Tactic | 82 |
Note
This is the first Agda dataset on Hugging Face. Agda is a dependently-typed programming language and proof assistant.
Use Cases
- Retrieval/RAG for Agda
- Learning dependent type patterns
- Training embeddings for formal proofs
- Cross-prover research (compare with Coq, Lean, Idris)