soco-ai / sf-qa Goto Github PK

Hey there,

In the cached results section, you put two links to the cached retrieval results of SPARTA retriever and BM25... but I found that these two links link to the same file...

I would like to ask that, in the paper SPARTA: Efficient Open-Domain Question Answering via Sparse Transformer Matching Retrieval... how many documents were you considering for the result table 1 about OpenSQuAD? i.e. top_k of retrieval

Thanks in advance for your assistance : )