APRIL: 258K Lean Proof-Repair Tuples with Compiler Diagnostics

The APRIL dataset contains 258,103 examples of Lean proof-repair tuples, including compiler diagnostics and explanations. It was created by the organization uw-math-ai and was last updated on the Hugging Face platform in February 2026. The dataset is intended for tasks related to formal verification and theorem proving.

Use Cases

• Train language models for automated proof repair based on the provided proof-repair tuples.
• Generate compiler error explanations based on the diagnostic information included in the dataset.
• Fine-tune models for joint proof repair and explanation generation, as referenced by the associated models gAPRIL-w-exp and gAPRIL-wo-exp.

Strengths

• Contains 258,103 examples, providing a substantial corpus for training.
• Includes compiler diagnostics and explanations for the proof-repair tuples.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but other metadata like file formats, license, and sample data is unavailable.

Provenance

Source

uw-math-ai on Hugging Face.

Collection Method

Likely extracted from Lean proof repositories and compiler outputs.

Time Range

null

Freshness

Last updated 2026-02-27 21:57:15; freshness should be verified.

Geography

null