Skip to content

Dataset · ikala

Taiwanese TMMLU+ benchmark across 66 subjects

ikala/tmmluplus

Multiple-choice questions in Traditional Chinese across 66 school and professional subjects for evaluating LLMs.

What's inside

The dataset contains question-answering tasks in Traditional Chinese covering 66 subjects ranging from school to professional levels. According to the card, it is about six times larger than TMMLU and better balanced; version v1.1 improves benchmark quality. The context largely reflects the linguistic, educational, and professional realities of Taiwan.

What you can do with it

  • Evaluating LLMs in Traditional Chinese
  • Comparing models across 66 subjects
  • Testing knowledge benchmarks
Tasks
Question answering
Languages
Chinese
Size
10K–100K rows
Downloads this month
6.8K
Likes
242

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("ikala/tmmluplus")