GEmO: Gemini-Empowered Olympiad Math Dataset

GEmO (Gemini-Empowered Olympiad Math Dataset) is a programmatically verified collection of advanced mathematical problem-solving trajectories. It contains exactly 11,500 unique rows of college-level, honors, and Olympiad-level mathematics across 11 specialized domains. The dataset was released by the Surpem team and was last updated on 2026-05-17.

Use Cases

• Training AI models for advanced mathematical problem-solving based on Olympiad-level problems.
• Benchmarking the reasoning capabilities of large language models based on rigorous, programmatically verified solution traces.
• Developing educational tools for college-level and honors mathematics based on the structured problem trajectories.
• Analyzing symbolic algebraic reasoning patterns based on the use of generators like SymPy mentioned in the description.

Strengths

• Contains exactly 11,500 unique rows of problem-solving data.
• Covers 11 specialized mathematical domains, indicating breadth.
• Problems and solutions are programmatically verified and constructed analytically, suggesting high rigor.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but other scale metrics like file size and formats are unknown.
• Freshness should be verified as the last update timestamp is in the future (2026-05-17).

Provenance

Source

Surpem team via Hugging Face.

Collection Method

Constructed analytically using symbolic algebraic generators (e.g., SymPy).

Freshness

Last updated 2026-05-17 18:17:15