Question about the search results

[Pangjing-Wu]

Hi there,

Hi, thank you for releasing this great benchmark! I'm currently running experiments using your dataset and had a couple of questions regarding the search_results files, such as dataset/search_results/arguana/sch_arguana_UAE-Large-V1_.txt.

From what I can tell, each file contains the top 1000 retrieved documents for each query. Could you please confirm if this understanding is correct? Additionally, I’m curious about the fifth column in these files, which appears to be a floating-point number. Could you clarify what this value represents and how it is calculated?

I have attached the head of the example file for your information.

1 Q0 training-health-hgwhwbutffs-con02b 0 0.5928019881248474 test
1 Q0 training-health-ahwba-pro03b 1 0.5631507635116577 test
1 Q0 training-health-mthwhwbpd-con02a 2 0.5619351267814636 test
1 Q0 validation-health-dhwiftj-pro02b 3 0.5603869557380676 test
1 Q0 training-health-hgwhwbutffs-con03b 4 0.5566678643226624 test
1 Q0 training-health-mthwhwbpd-pro02b 5 0.5546292066574097 test

Thank you in advance!