Competition-Level Math Problems With Grading Rubrics

March 2026 dataset contains 50 challenging mathematics reasoning problems. Each problem includes a detailed reference solution, a structured grading rubric, and an anonymized model evaluation attempt.

Use Cases

• Benchmark model performance on multi-step reasoning problems using the structured grading rubric for evaluation.
• Analyze frontier model attempts to identify common failure patterns in proof construction or complex computation.
• Train or fine-tune models for mathematical text generation using the detailed reference solutions as supervision.
• Develop automated grading systems by leveraging the point-by-point grading criteria provided in the rubrics.

Strengths

• Contains 50 competition-level problems designed for high difficulty.
• Each entry includes a structured rubric with point-by-point grading criteria.
• Provides anonymized model evaluation results against the rubrics for benchmarking.

Limitations

• Small sample size of only 50 problems limits statistical power for training.
• Limited to English language problems, restricting multilingual applicability.
• Focus on competition-level math may not represent broader mathematical reasoning tasks.

Provenance

Source

BakeAI

Collection Method

Collection of challenging mathematics problems with constructed rubrics and model evaluations.

Time Range

null

Freshness

Last updated March 2026.

Geography

null