This dataset contains the Turkish translations of problems from the 2025 American Invitational Mathematics Examination (AIME). It is intended to serve as a benchmark for evaluating the advanced mathematical reasoning capabilities of Large Language Models (LLMs) in the Turkish language.
The questions were translated into Turkish using GPT-5, then manually verified and corrected. Additional quality checks were performed to identify formatting issues, LaTeX… See the full description on the dataset page:
https://huggingface.co/datasets/ytu-ce-cosmos/aime25-tr.
