README.md

---
license: apache-2.0
configs:
- config_name: aidayago2
  data_files:
  - split: train
    path: aidayago2/train-*
  - split: validation
    path: aidayago2/validation-*
  - split: test
    path: aidayago2/test-*
- config_name: blink
  data_files:
  - split: train
    path: blink/train-*
- config_name: cweb
  data_files:
  - split: validation
    path: cweb/validation-*
  - split: test
    path: cweb/test-*
- config_name: eli5
  data_files:
  - split: train
    path: eli5/train-*
  - split: validation
    path: eli5/validation-*
  - split: test
    path: eli5/test-*
- config_name: fever
  data_files:
  - split: train
    path: fever/train-*
  - split: validation
    path: fever/validation-*
  - split: test
    path: fever/test-*
- config_name: hotpotqa
  data_files:
  - split: train
    path: hotpotqa/train-*
  - split: validation
    path: hotpotqa/validation-*
  - split: test
    path: hotpotqa/test-*
- config_name: nq
  data_files:
  - split: train
    path: nq/train-*
  - split: validation
    path: nq/validation-*
  - split: test
    path: nq/test-*
- config_name: structured_zeroshot
  data_files:
  - split: train
    path: structured_zeroshot/train-*
  - split: validation
    path: structured_zeroshot/validation-*
  - split: test
    path: structured_zeroshot/test-*
- config_name: trex
  data_files:
  - split: train
    path: trex/train-*
  - split: validation
    path: trex/validation-*
  - split: test
    path: trex/test-*
- config_name: triviaqa
  data_files:
  - split: train
    path: triviaqa/train-*
  - split: validation
    path: triviaqa/validation-*
  - split: test
    path: triviaqa/test-*
- config_name: wned
  data_files:
  - split: validation
    path: wned/validation-*
  - split: test
    path: wned/test-*
- config_name: wow
  data_files:
  - split: train
    path: wow/train-*
  - split: validation
    path: wow/validation-*
  - split: test
    path: wow/test-*
dataset_info:
- config_name: aidayago2
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: train
    num_bytes: 55864609
    num_examples: 18395
  - name: validation
    num_bytes: 15955325
    num_examples: 4784
  - name: test
    num_bytes: 12471688
    num_examples: 4463
  download_size: 18666928
  dataset_size: 84291622
- config_name: blink
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: train
    num_bytes: 175321157
    num_examples: 100000
  download_size: 111935980
  dataset_size: 175321157
- config_name: cweb
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: validation
    num_bytes: 51475085
    num_examples: 5599
  - name: test
    num_bytes: 56279580
    num_examples: 5543
  download_size: 23161761
  dataset_size: 107754665
- config_name: eli5
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: train
    num_bytes: 609110956
    num_examples: 200000
  - name: validation
    num_bytes: 12214766
    num_examples: 1507
  - name: test
    num_bytes: 769699
    num_examples: 600
  download_size: 381533571
  dataset_size: 622095421
- config_name: fever
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: train
    num_bytes: 130671464
    num_examples: 104966
  - name: validation
    num_bytes: 13054824
    num_examples: 10444
  - name: test
    num_bytes: 12535921
    num_examples: 10100
  download_size: 86214810
  dataset_size: 156262209
- config_name: hotpotqa
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: train
    num_bytes: 118290457
    num_examples: 88869
  - name: validation
    num_bytes: 7385175
    num_examples: 5600
  - name: test
    num_bytes: 7236506
    num_examples: 5569
  download_size: 92994403
  dataset_size: 132912138
- config_name: nq
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: train
    num_bytes: 111181156
    num_examples: 87372
  - name: validation
    num_bytes: 3699165
    num_examples: 2837
  - name: test
    num_bytes: 1805683
    num_examples: 1444
  download_size: 73218512
  dataset_size: 116686004
- config_name: structured_zeroshot
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: train
    num_bytes: 188554112
    num_examples: 147909
  - name: validation
    num_bytes: 4754817
    num_examples: 3724
  - name: test
    num_bytes: 6242056
    num_examples: 4966
  download_size: 128366579
  dataset_size: 199550985
- config_name: trex
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: train
    num_bytes: 267341864
    num_examples: 200000
  - name: validation
    num_bytes: 6672336
    num_examples: 5000
  - name: test
    num_bytes: 6292578
    num_examples: 5000
  download_size: 205161484
  dataset_size: 280306778
- config_name: triviaqa
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: train
    num_bytes: 112256098
    num_examples: 61817
  - name: validation
    num_bytes: 9990772
    num_examples: 5355
  - name: test
    num_bytes: 8343733
    num_examples: 6578
  download_size: 75758558
  dataset_size: 130590603
- config_name: wned
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: validation
    num_bytes: 10296438
    num_examples: 3396
  - name: test
    num_bytes: 10594837
    num_examples: 3376
  download_size: 6055450
  dataset_size: 20891275
- config_name: wow
  features:
  - name: query
    dtype: string
  - name: query_id
    dtype: string
  - name: answers
    sequence: string
  - name: context_doc_ids
    sequence: string
  splits:
  - name: train
    num_bytes: 108554875
    num_examples: 63734
  - name: validation
    num_bytes: 5183194
    num_examples: 3054
  - name: test
    num_bytes: 4704307
    num_examples: 2944
  download_size: 63159586
  dataset_size: 118442376
---

# KILT Benchmark with Top-k Retrieval Results

This dataset is a modified version of the [KILT Benchmark](https://github.com/facebookresearch/KILT) from the paper "KILT: a benchmark for knowledge intensive language tasks". It includes additional top-k retrieval results used in the paper "[Chain-of-Retrieval Augmented Generation](https://arxiv.org/pdf/2501.14342)".

## Differences from the Original KILT Dataset

The primary difference is the addition of the `context_doc_ids` field. This field provides the IDs of the top-k documents retrieved during the CoRAG experiments. You can use these IDs to retrieve the corresponding document content from the `corag/kilt-corpus` dataset at [https://huggingface.co/datasets/corag/kilt-corpus](https://huggingface.co/datasets/corag/kilt-corpus).

## Fields

- `query_id`: The ID of the query.
- `query`: The text of the query.
- `answers`: The answers to the query.
- `context_doc_ids`: The IDs of the top-k documents retrieved for the query.

## Usage

To use this dataset, you can load it using the Hugging Face `datasets` library:

```python
from datasets import load_dataset

dataset = load_dataset("corag/kilt", "hotpotqa", split="train")

print(dataset)
print(dataset[0])
```

## Caveats

1. Due to issues with ID mapping, a small number of questions are missing from the TriviaQA dataset. However, this is expected to have a negligible impact on overall evaluation metrics.

## References

```
@inproceedings{petroni2021kilt,
  title={KILT: a benchmark for knowledge intensive language tasks},
  author={Petroni, Fabio and Lewis, Patrick and Rockt{\"a}schel, Tim and Riedel, Sebastian and Yih, Wen-tau},
  booktitle={Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies},
  pages={1113--1131},
  year={2021}
}

@article{wang2025chain,
  title={Chain-of-Retrieval Augmented Generation},
  author={Wang, Liang and Chen, Haonan and Yang, Nan and Huang, Xiaolong and Dou, Zhicheng and Wei, Furu},
  journal={arXiv preprint arXiv:2501.14342},
  year={2025}
}
```