This dataset is a Nano-style Brazilian Portuguese retrieval benchmark for HAKARI-Bench. It is derived from the retrieval and reranking tasks exposed by MTEB-BR at commit e17f36ae304e7bdb76202744c5fb56b440050213. The six public splits are Quati, JurisTCU, BRTaxQAR, FaQuADIR, MedPTRetrieval, and FaqBacenRetrieval.

Usage

from datasets import load_dataset

dataset_id = "hakari-bench/NanoMTEB-BR" split = "MedPTRetrieval"

queries = load_dataset(dataset_id… See the full description on the dataset page: https://huggingface.co/datasets/hakari-bench/NanoMTEB-BR.