Commit
·
8e196b5
1
Parent(s):
df57a76
feat: updated README
Browse files
README.md
CHANGED
@@ -1,15 +1,15 @@
|
|
1 |
---
|
2 |
configs:
|
3 |
-
- config_name:
|
4 |
-
data_files: crag_task_1_and_2/
|
5 |
-
- config_name:
|
6 |
-
data_files: crag_task_1_and_2/
|
7 |
-
- config_name:
|
8 |
-
data_files: crag_task_1_and_2/
|
9 |
-
- config_name:
|
10 |
-
data_files: crag_task_1_and_2/
|
11 |
-
- config_name:
|
12 |
-
data_files: crag_task_1_and_2/
|
13 |
license: cc-by-nc-4.0
|
14 |
task_categories:
|
15 |
- question-answering
|
@@ -26,6 +26,30 @@ pretty_name: CRA
|
|
26 |
|
27 |
Datasets are taken from Facebook's [CRAG: Comprehensive RAG Benchmark](https://github.com/facebookresearch/CRAG), see their [arXiv paper](https://arxiv.org/abs/2406.04744) for details about the dataset construction.
|
28 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
CRAG (Comprehensive RAG Benchmark) is a rich and comprehensive factual question answering benchmark designed to advance research in RAG. The public version of the dataset includes:
|
30 |
- 2706 Question-Answer pairs
|
31 |
- 5 domains: Finance, Sports, Music, Movie, and Open domain
|
@@ -34,7 +58,7 @@ CRAG (Comprehensive RAG Benchmark) is a rich and comprehensive factual question
|
|
34 |

|
35 |
|
36 |
The datasets `crag_task_1_and_2_dev_v4_subsample_*.json.bz2` have been created from the dataset [crag_task_1_and_2_dev_v4.jsonl.bz2](https://github.com/facebookresearch/CRAG/raw/refs/heads/main/data/crag_task_1_and_2_dev_v4.jsonl.bz2?download=) available on CRAG's GitHub repository.
|
37 |
-
For an easier handling and download of the dataset, we have used
|
38 |
1. We have created a new label `answer_type`, classifying the answers in 3 categories:
|
39 |
- `invalid` for any answer == "invalid question"
|
40 |
- `no_answer` for any answer == "i don't know"
|
|
|
1 |
---
|
2 |
configs:
|
3 |
+
- config_name: crag_task_1_and_2_subset_1
|
4 |
+
data_files: crag_task_1_and_2/crag_task_1_and_2_dev_v4_subset_cdf90925.jsonl.bz2
|
5 |
+
- config_name: crag_task_1_and_2_subset_2
|
6 |
+
data_files: crag_task_1_and_2/crag_task_1_and_2_dev_v4_subset_2e617280.jsonl.bz2
|
7 |
+
- config_name: crag_task_1_and_2_subset_3
|
8 |
+
data_files: crag_task_1_and_2/crag_task_1_and_2_dev_v4_subset_9bca9871.jsonl.bz2
|
9 |
+
- config_name: crag_task_1_and_2_subset_4
|
10 |
+
data_files: crag_task_1_and_2/crag_task_1_and_2_dev_v4_subset_ff5aa423.jsonl.bz2
|
11 |
+
- config_name: crag_task_1_and_2_subset_5
|
12 |
+
data_files: crag_task_1_and_2/crag_task_1_and_2_dev_v4_subset_3f4353ba.jsonl.bz2
|
13 |
license: cc-by-nc-4.0
|
14 |
task_categories:
|
15 |
- question-answering
|
|
|
26 |
|
27 |
Datasets are taken from Facebook's [CRAG: Comprehensive RAG Benchmark](https://github.com/facebookresearch/CRAG), see their [arXiv paper](https://arxiv.org/abs/2406.04744) for details about the dataset construction.
|
28 |
|
29 |
+
# CRAG Sampler
|
30 |
+
|
31 |
+
We have added a simple Python tool for performing stratified sampling on CRAG data.
|
32 |
+
|
33 |
+
## Installation
|
34 |
+
|
35 |
+
### Local Development Install (Recommended)
|
36 |
+
```bash
|
37 |
+
git clone https://huggingface.co/Quivr/CRAG.git
|
38 |
+
cd CRAG
|
39 |
+
pip install -r requirements.txt # Install dependencies
|
40 |
+
pip install -e . # Install package in development mode
|
41 |
+
```
|
42 |
+
|
43 |
+
## Quick Start
|
44 |
+
|
45 |
+
### Running the example
|
46 |
+
|
47 |
+
```bash
|
48 |
+
python -m examples.basic_sampling
|
49 |
+
```
|
50 |
+
|
51 |
+
# CRAG dataset
|
52 |
+
|
53 |
CRAG (Comprehensive RAG Benchmark) is a rich and comprehensive factual question answering benchmark designed to advance research in RAG. The public version of the dataset includes:
|
54 |
- 2706 Question-Answer pairs
|
55 |
- 5 domains: Finance, Sports, Music, Movie, and Open domain
|
|
|
58 |

|
59 |
|
60 |
The datasets `crag_task_1_and_2_dev_v4_subsample_*.json.bz2` have been created from the dataset [crag_task_1_and_2_dev_v4.jsonl.bz2](https://github.com/facebookresearch/CRAG/raw/refs/heads/main/data/crag_task_1_and_2_dev_v4.jsonl.bz2?download=) available on CRAG's GitHub repository.
|
61 |
+
For an easier handling and download of the dataset, we have used our CRAG sampler to split the 2706 rows of the original file in 5 subsamples, following the procedure below:
|
62 |
1. We have created a new label `answer_type`, classifying the answers in 3 categories:
|
63 |
- `invalid` for any answer == "invalid question"
|
64 |
- `no_answer` for any answer == "i don't know"
|