The goal of this project is to create a high-quality (golden) dataset from your data, which will serve as a benchmark for evaluating various embedding models.

Korean-optimized RAG evaluation toolkit with Kiwi tokenizer, ROUGE metrics, and IR evaluation for retrieval systems (Hit@K, NDCG@K, MRR, etc.)