Dataset · ginigen-ai
Benchmark for evaluating compact language models
ginigen-ai/smol-worldcup
A multilingual dataset for comprehensive testing of small LLMs across five axes: from accuracy to efficiency.
What's inside
A small set (up to 1000 examples) in text and table formats across six languages: English, Korean, Arabic, Portuguese, Turkish, and Bengali. The dataset was created for the Smol AI WorldCup project—a framework for evaluating compact language models based on five criteria, including fairness, speed, size, and efficiency.
What you can do with it
- Benchmarking and comparing compact language models across aggregate quality, speed, and size
- Testing multilingual capabilities of small LLMs in text generation and QA tasks
- Tasks
- Text generation, Question answering
- Languages
- English, Korean, Arabic, Portuguese, Turkish, Bangla, Thai
- Size
- up to 1K rows
- Downloads this month
- 402
- Likes
- 47
How to load it
from datasets import load_dataset
ds = load_dataset("ginigen-ai/smol-worldcup")