Dataset Card for Thai_Exam

ThaiExam is a Thai knowledge benchmarking dataset, consisting of multiple-choice questions from examinations in Thailand. The dataset was originally developed for evaluating Typhoon (Thai LLM). This dataset contains 5 splits corresponding to 5 examinations as follows:

1. ONET: The Ordinary National Educational Test (ONET) is an examination for students in Thailand. This dataset is based on the grade-12 ONET exam, comprising 4 subjects and each question has 5 choices. These subjects are Thai, Mathematics, Social Studies, and Science, amounting to a total of 167 questions and options.

2. IC: The Investment Consultant (IC) examination is a licensing test for investment professionals in Thailand. The exam questions were developed by the Stock Exchange of Thailand (SET), featuring 4 choices per question. We extracted questions for levels 1, 2, and 3 resulting in a total of 100 questions and options.

3. TGAT: The Thai General Aptitude Test (TGAT) is a national high school examination in Thailand. The exam focuses on critical and logical thinking skills. We collected a total of 70 questions and answers where each question consists of 4 choices per question.

4. TPAT-1: The Thai Professional Aptitude Test 1 (TPAT-1) is a national high school examination in Thailand. The Exam assesses students’ professional skills requirement in medical schools. This subset contains reasoning and medical ethics. We collected a total of 121 questions and answers. The TPAT-1 consists of 5 choices per question.

5. A-Level: The A-Level exam is an academic knowledge assessment examination (Applied Knowledge Level) covering general foundational subjects. The content assessed in this examination aligns with the curriculum guidelines and emphasizes the practical application of knowledge in daily life. We collected a total of 132 questions and answers.

Usage

For each exam, the train split, consisting of 5 examples, can be used for 5-shot evaluation (in-context examples), and we provide a 5-shot evaluation script here.

Version

This dataset was initially developed for evaluating Typhoon. As we made improvements and fixed errors in the dataset, here are the versions:

• v0.1: As per results reported in Typhoon technical report and Typhoon-1.5 Blog. This pre-release version can be downloaded from this link.
• v1.0: The current version (revision: d78aef04ea3cc5095545e6951cb39e17c64e26a1)

HELM

This dataset has been integrated into the HELM leaderboard by Stanford CRFM.

Citation Information

This dataset was curated by the Typhoon team, and the first iteration of this dataset was used in the evaluation of Typhoon. If you find this dataset useful, you can cite our work:

@article{pipatanakul2023typhoon,
  title={Typhoon: Thai Large Language Models},
  author={Pipatanakul, Kunat and Jirabovonvisut, Phatrasek and Manakul, Potsawee and Sripaisarnmongkol, Sittipong and Patomwong, Ruangsak and Chokchainant, Pathomporn and Tharnpipitchai, Kasima},
  journal={arXiv preprint arXiv:2312.13951},
  year={2023}
}