AI4Bharat - Hercule-BN: Bengali Language Translation Evaluation Model

About Model

Hercule-bn is a Bengali-language evaluation model from the CIA Suite, designed to assess multilingual Large Language Models (LLMs). By leveraging English reference responses, it evaluates Bengali outputs, ensuring accurate scoring and alignment with human evaluations. The model is fine-tuned on the INTEL dataset and supports zero-shot evaluations for languages not seen during training. It provides feedback with scores ranging from 1 to 5, and users can access wrapper functions and classes for seamless integration from the associated GitHub repository.