---
dataset_info:
  features:
  - name: id
    dtype: int64
  - name: question
    dtype: string
  - name: pred_answer
    dtype: string
  - name: topk
    struct:
    - name: contents
      sequence: string
    - name: docid
      sequence: int64
  - name: answer
    sequence: string
  splits:
  - name: train
    num_bytes: 1244684721
    num_examples: 222849
  download_size: 706547692
  dataset_size: 1244684721
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

The dataset is originated by [dmrau/multi_qa](dmrau/multi_qa).

The *topk* passages are retrieved by SPLADE-v3 and the *pred_answer* is generated with Mistral-7B-Instruct-v0.2 with the retrieved passages.