This repository contains the released version of the SCALE dataset used in the paper "Scaling Cultural Resources for Improving Generative Models".
SCALE is a general-purpose cultural knowledge dataset designed to support analysis of how generative models represent cultures and countries. The dataset consists of culturally salient artifacts associated with individual countries across multiple aspects, such as:
- Cuisine
- Holidays and Festivals
- Clothing and Accessories
- Landmarks
- Historical Events
- Sportspeople
- Sports Teams
The current release covers 29 countries.