dataset.py

from datasets import DatasetDict, load_dataset

def load_nc_bench():
    data_files = {
        "basic": "basic/test.jsonl",
    	"rag": "rag/test.jsonl",
    	"complex_request": "complex_request/test.jsonl",
    }
    return DatasetDict({
        split: load_dataset("json", data_files={split: path})[split]
        for split, path in data_files.items()
    })