JMTEB is a benchmark for evaluating Japanese text embedding models. It consists of 5 tasks, currently involving 28 datasets in total. You can find the update history here.

TL;DR

from datasets import load_dataset

dataset = load_dataset("sbintuitions/JMTEB", name="<dataset_name>", split="")

JMTEB_DATASET_NAMES = ( 'livedoor_news', 'mewsc16_ja', 'sib200_japanese_clustering'… See the full description on the dataset page: https://huggingface.co/datasets/sbintuitions/JMTEB.