Crosslingual Rag

This project explores Cross-Lingual Retrieval-Augmented QA with a focus on handling Hindi questions within English contexts. Key stages include translation, evidence retrieval, and QA generation, evaluated using metrics like Recall@K, EM, and F1. It leverages datasets like XQuAD and FLORES-200 while analyzing translation accuracy (e.g., BLEU 40). Baseline results highlight retrieval and QA challenges, with proposed improvements such as enhanced chunking and scaling models. Error analysis...