ThaiExam
SCB 10XThaiExam is a Thai knowledge benchmark of multiple-choice questions taken from real examinations in Thailand: the grade-12 O-NET, the Investment Consultant licensing exam, TGAT, TPAT-1 and A-Level. It holds 590 questions in total, and each exam includes five worked examples for few-shot prompting. The Typhoon team built it to evaluate the Typhoon Thai language models.
Openness
5 high confidence- license
- apache-2.0(card metadata)
- access
- public(ungated, answers included)
- dataset_card
- present(per-exam description, versions and HELM note)
The questions and answers download without a gate under Apache-2.0, and the card describes each exam and the dataset's revisions. Nothing is held back for private scoring.
- https://huggingface.co/api/datasets/typhoon-ai/thai_exam recorded 2026-09-24
gated: false; license:apache-2.0
- https://huggingface.co/datasets/typhoon-ai/thai_exam/raw/main/README.md recorded 2026-09-24
license: apache-2.0; five exam splits with train (5 examples) and test files
Adoption
2 high confidenceAdoption is Hugging Face downloads of the repository. Leaderboards that bundle the test, such as HELM and SEA-HELM, may run it from their own copies, which this count does not see.
- https://huggingface.co/api/datasets/typhoon-ai/thai_exam recorded 2026-09-24
2167 downloads in the trailing 30 days for typhoon-ai/thai_exam
Capability
4 medium confidenceThaiExam uses real Thai exam questions and is built into Stanford's HELM Thai leaderboard, the Thai LLM leaderboard and SEA-HELM. It is a single-language set of 590 questions, and SEA-HELM, which includes it, covers many more tasks and languages.
- https://arxiv.org/abs/2312.13951 recorded 2026-09-24
"we develop ThaiExam, a benchmark based on examinations for high-school students and investment professionals in Thailand"
- https://huggingface.co/datasets/typhoon-ai/thai_exam recorded 2026-09-24
Number of rows: 590; Space using typhoon-ai/thai_exam: ThaiLLM-Leaderboard/leaderboard
- https://huggingface.co/datasets/typhoon-ai/thai_exam/raw/main/README.md recorded 2026-09-24
"This dataset has been integrated into the HELM leaderboard by Stanford CRFM"; ONET 167, IC 100, TGAT 70, TPAT-1 121, A-Level 132 questions
- https://raw.githubusercontent.com/aisingapore/SEA-HELM/main/docs/datasets_and_prompts.md recorded 2026-09-24
Knowledge table: "Thai | Thai Exam | Native | General | Apache 2.0"
Verified 2026-09-24