RAG Evaluation Scripts

[qiaoruiyt]

Hi authors!

Thank you for curating the nice dataset! May I ask whether it is possible to release the RAG evaluation scripts used to generate "Table 4: Question-answering results with different retrievers." in v3 of the paper on Arxiv? We are trying to better understand how IR improves RAG subsequently. Thank you!

[Frankgu3528]

I tried to implement the RAG generation and evaluation process on my own and found the “None” results were almost always better then results with retrieved document. I tried Llama-3.1-70B-Instruct and Gemini-2.0 as generation and evaluation model, tried different retrievers but the results were the same. I would greatly appreciate it if you could share the RAG evaluation scripts. Thanks in advance!