This dataset is a Nano-style retrieval dataset for HAKARI-bench. NanoLongEmbed contains Nano-style long-context retrieval splits derived from LongEmbed tasks.

Usage

from datasets import load_dataset

dataset_id = "hakari-bench/NanoLongEmbed" split = "Nano2WikiMultihopQA"

queries = load_dataset(dataset_id, "queries", split=split) corpus = load_dataset(dataset_id, "corpus", split=split) qrels = load_dataset(dataset_id, "qrels", split=split)… See the full description on the dataset page: https://huggingface.co/datasets/hakari-bench/NanoLongEmbed.