Skip to content

Dataset · CohereLabsCommunity

Cultural riddles in 51 languages by native speakers from Cohere Labs

CohereLabsCommunity/CulturalRiddles

Multilingual benchmark of original cultural riddles written by native speakers from 62 communities and 45 countries.

What's inside

This is a community-created benchmark based on tabular and text data (100k to 1M records). It features original cultural riddles in 51 languages (including rare Arabic dialects and Assamese), written by 167 native speakers from 62 communities and 45 countries. The project was prepared with the participation of the Cohere Labs Community to evaluate models' cultural awareness.

What you can do with it

  • Evaluating cultural knowledge and reasoning capabilities of language models on low-resource languages and dialects
  • Testing multilingual models for understanding local context and allegories
Tasks
Question answering
Languages
Tunisian Arabic, apc, apd, Moroccan Arabic, Egyptian Arabic, Assamese, Belarusian, Bangla and 4 more
Size
100K–1M rows
Downloads this month
827
Likes
45

How to load it

from datasets import load_dataset

ds = load_dataset("CohereLabsCommunity/CulturalRiddles")