MIRACL

The MIRACL dataset is a multilingual benchmark that contains native-language questions and responses generated by large language models, used to evaluate the faithfulness and relevance of retrieval augmented generation systems as assessed by expert annotators.

Papers using MIRACL (3)

LLMs Encode Relevance as a Layer-Wise Cross-Lingual Signal2026KaLM-Reranker-V1: Fast but Not Late Interaction for Compressed Document Reranking2026MEMERAG: A Multilingual End-to-End Meta-Evaluation Benchmark for Retrieval Augmented Generation2025