tubitak-science-olympiad-tr

TUBITAK科学奥赛数据集包含来自土耳其科学技术研究委员会（TUBITAK）科学奥赛的多选题和开放式科学问题，旨在作为评估大型语言模型（LLMs）在土耳其语言中高级分析、数学和计算推理能力的基准。数据集涵盖约2700个问题，分布在五个领域：计算机科学、物理、数学、中学计算机科学和中学数学。每个问题条目包含唯一标识符、学科领域、年份、考试阶段、问题编号、问题图像、解答图像（如有）、LaTeX格式的问题和解答文本、以及多个布尔标志和选项值。数据集特别适用于测试LLMs在非英语环境下的多步科学推理任务、多模态评估以及链式思维（CoT）能力。数据集在CC BY 4.0许可下发布，适用于研究和教育目的。

The TUBITAK Science Olympiad Dataset contains multiple-choice and open-ended scientific questions sourced from the competitions hosted by the Turkish Scientific and Technological Research Council (TUBITAK). It is designed as a benchmark for evaluating the advanced analytical, mathematical, and computational reasoning capabilities of Large Language Models (LLMs) in Turkish. The dataset covers approximately 2,700 questions distributed across five domains: Computer Science, Physics, Mathematics, Secondary School Computer Science, and Secondary School Mathematics. Each question entry includes a unique identifier, subject domain, year, exam stage, question number, question image, solution image (if available), LaTeX-formatted question and solution texts, as well as multiple boolean flags and option values. This dataset is particularly well-suited for testing LLMs on multi-step scientific reasoning tasks, multimodal evaluations, and Chain-of-Thought (CoT) capabilities in non-English environments. The dataset is released under the CC BY 4.0 license and is intended for research and educational purposes.