← all datasets

CompMix-IR

Emerging
1papers using it
2024first seen

CompMix-IR is a heterogeneous knowledge retrieval benchmark that includes two retrieval scenarios with various instructions, comprising over 9,400 question-answer pairs and a corpus of 10 million entries across four different types of data, used to evaluate the performance of information retrieval models.

🔎 Find this dataset

Papers using CompMix-IR (1)

CompMix-IR — datasets — learning-to-hash