AI4Bharat - Hercule-HI: Hindi language translation Evaluation Model

About Model

Hercule-hi is a Hindi-language evaluation model from the CIA Suite, aimed at assessing multilingual Large Language Models (LLMs). It evaluates Hindi outputs by comparing them to English reference responses, ensuring accurate scoring and better alignment with human judgments. Fine-tuned on the INTEL dataset, the model supports zero-shot evaluations on unseen languages and provides feedback with scores from 1 to 5. Users can access wrapper functions and classes for seamless integration from the associated GitHub repository.