Formal Conjectures: 4,557 Lean 4 Statements of Open Problems and Conjectures (Google DeepMind formal-conjectures, phanerozoic extraction)

Provider
Google DeepMind and formal-conjectures contributors; extraction by phanerozoic
License
Apache-2.0 (source)
Size
3 files
Format
Parquet (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/phanerozoic/Lean4-FormalConjectures)
Access
Free and public. No account with the data provider is needed.
Category
Mathematics
Source
https://github.com/google-deepmind/formal-conjectures

About this dataset

Declarations from Google DeepMind's formal-conjectures repository, a collection of Lean 4 formalizations of open mathematical conjectures and problems, such as Erdős problems, together with solved variants and supporting lemmas. There are 4,557 entries from 896 files in 88 sub-libraries, extracted at upstream commit b9b8aa0f. 3,759 entries carry docstrings with the informal problem statement and source. Each row gives the formal statement, the proof body (often sorry for statements that are still open), the declaration kind, the name, the library and file, the imports, and the identifiers it uses. The repository serves as a benchmark target for automated theorem provers and for checking autoformalization against human-written formal statements of research-level problems. It is one Parquet file; load it with pandas or Hugging Face datasets. The upstream formal-conjectures repository is released by Google DeepMind under the Apache-2.0 license.

How to load this dataset in Python

from datasets import load_dataset
ds = load_dataset("phanerozoic/Lean4-FormalConjectures")
print(ds)

Data structure

File types in sampled files: Parquet

Files

Example file: data/train-00000-of-00001.parquet

Analyze this dataset without downloading it

On Scigantic, Formal Conjectures: 4,557 Lean 4 Statements of Open Problems and Conjectures (Google DeepMind formal-conjectures, phanerozoic extraction) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Mathematics datasets