The aim of the project is to develop a state-of-the-art Indian Cultural benchmark that can test these models for their cultural accuracies, especially in a country like India which is rich in diversity. With Hugging Face support, our aim is to expand the benchmark into an alignment dataset and release it.
We will be covering the 15 official languages of India: Assamese, Bengali, Gujarati, Hindi, Kannada, Kashmiri, Konkani, Malayalam, Manipuri, Marathi, Odia, Punjabi, Sanskrit, Tamil, Telugu and their corresponding dialects. This is a community-driven project and your support and collaboration can make this initiative a success.
Project is under development. More coming soon.