Datasets:

Modalities:
Text
Formats:
parquet
Languages:
English
ArXiv:
Libraries:
Datasets
Dask

Dataset Inconsistency

#7
by koleckar - opened

Hello,

we have found count severe mismatch between mteb and microsoft msmarco.

queries: ~1M in microsoft/msmarco-v2 , 285k in mteb/msmarco-v2
corpus: ~1M * ~10 passages per query in microsoft/msmarco-v2, 133M in mteb/msmarco-v2

Is the v2.1 here the same or is it some other dataset?

Thank you,

w/regards
D

Sign up or log in to comment