QO-Bench benchmark evaluates query-operator-preserving retrieval in RAG

Arxiv CS.CL (NLP/language models)

The benchmark tests RAG systems on natural-language versions of database-style queries over typed event tuples. It focuses on preserving query operators during retrieval.

ragbenchmarksretrieval

Arxiv CS.CL (NLP/language models)·Mengao ZhangXiang YangChang LiuTianhui TanKe-wei Huang

7 days ago