IMO-AnswerBench: 400 Challenging Math Olympiad Problems

IMO-AnswerBench is a benchmark dataset for evaluating mathematical reasoning in large language models. It consists of 400 challenging short-answer problems from the International Mathematical Olympiad and other sources. The dataset was released by Google DeepMind in conjunction with their 2025 IMO gold medal achievement and is hosted by OpenEvals.

Use Cases

• Benchmarking AI model performance on complex mathematical reasoning based on the described 400 problems.
• Training or fine-tuning models for advanced mathematical problem-solving based on the Olympiad-level content.
• Analyzing the types of mathematical challenges AI models struggle with based on the curated short-answer format.

Strengths

• Contains 400 specific problems for evaluation.
• Problems are sourced from the International Mathematical Olympiad, a high-standard competition.
• Released in conjunction with a documented AI achievement (Google DeepMind's 2025 IMO gold).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

International Mathematical Olympiad (IMO) and other sources, released by Google DeepMind.

Collection Method

Likely curated and compiled for benchmark purposes.

Freshness

Last updated 2026-01-23 16:26:25; freshness should be verified.