Dataset · di-zhang-fdu
AIME Math Competition Benchmark (1983-2024)
di-zhang-fdu/AIME_1983_2024
Collection of AIME math competition problems from 1983-2024 for unbiased LLM reasoning evaluation.
What's inside
This dataset contains under 1,000 records in tabular and text format, collected from the Art of Problem Solving website. It features AIME math competition problems from 1983 through 2023, along with the second part of 2024. The authors emphasize that this is a strict test benchmark and must not be used for model training.
What you can do with it
- Evaluating mathematical reasoning and logic capabilities of language models
- Comparing prompts and agents on challenging competition problems
- Size
- up to 1K rows
- Downloads this month
- 24.9K
- Likes
- 41
How to load it
from datasets import load_dataset
ds = load_dataset("di-zhang-fdu/AIME_1983_2024")