The FATE (Formal Algebra Theorem Evaluation) benchmarks.
This collection contains three benchmarks in abstract algebra and commutative algebra:
- FATE-M (150 exercises)
- FATE-H (100 exercises)
- FATE-X (100 exercises)
All exercises are formalized in Lean.
The problems in these benchmarks are sourced from:
- Undergraduate and graduate textbooks,
- Final exams and PhD qualifying exams,
- Research literature.
Broadly speaking, the three benchmarks have progressively increasing difficulty:
- FATE-M consists primarily of standard textbook exercises (basic level),
- FATE-H contains challenging problems from abstract/commutative algebra final exams,
- FATE-X includes problems at PhD qualifying exam level and beyond.
Each Lean file in every benchmark contains:
- One fully formalized statement,
- A single
sorry, - Appropriate open namespaces at the beginning,
- Natural language annotations preceding the statement.
Key differences:
- FATE-M and FATE-H: No extra definitions are included,
- FATE-X: Minimal dependent definitions formalized before the statement.
For users' convenience, we provide a PDF file and a JSON file for each benchmark to make reading and usage easier.
We strongly recommend AGAINST combining these benchmarks due to their:
- Significantly different difficulty levels,
- Distinct formalization characteristics.
The FATE-M benchmark is a refactored version of the benchmark referenced in:
REAL-Prover: Retrieval-Augmented Lean Prover for Mathematical Reasoning