updated readme
Browse files
README.md
CHANGED
|
@@ -18,6 +18,7 @@ language:
|
|
| 18 |
- te
|
| 19 |
- th
|
| 20 |
- zh
|
|
|
|
| 21 |
multilinguality:
|
| 22 |
- multilingual
|
| 23 |
pretty_name: NoMIRACL
|
|
@@ -31,7 +32,7 @@ license:
|
|
| 31 |
- apache-2.0
|
| 32 |
---
|
| 33 |
|
| 34 |
-
# Dataset Card for NoMIRACL
|
| 35 |
|
| 36 |
Retrieval Augmented Generation (RAG) is a powerful approach to incorporate external knowledge into large language models (LLMs) to enhance the accuracy and faithfulness of generated responses. However, evaluating LLM robustness in RAG across different language families has been a challenge, leading to gaps in understanding the model's performance against errors in external retrieved knowledge. To address this, we present NoMIRACL, a human-annotated dataset designed for evaluating LLM robustness in RAG across 18 diverse languages.
|
| 37 |
|
|
@@ -65,8 +66,8 @@ nomiracl = datasets.load_dataset('miracl/nomiracl', language, split=f'{split}.{s
|
|
| 65 |
Under folders `data/{lang}`,
|
| 66 |
the subset of corpus is saved in `.jsonl.gz` format, with each line to be:
|
| 67 |
```
|
| 68 |
-
{"docid": "28742#27",
|
| 69 |
-
"title": "Supercontinent",
|
| 70 |
"text": "Oxygen levels of the Archaean Eon were negligible and today they are roughly 21 percent. [ ... ]"}
|
| 71 |
```
|
| 72 |
|
|
@@ -99,14 +100,14 @@ for data in nomiracl: # or 'dev', 'testA'
|
|
| 99 |
query = data['query']
|
| 100 |
positive_passages = data['positive_passages']
|
| 101 |
negative_passages = data['negative_passages']
|
| 102 |
-
|
| 103 |
for entry in positive_passages: # OR 'negative_passages'
|
| 104 |
docid = entry['docid']
|
| 105 |
title = entry['title']
|
| 106 |
text = entry['text']
|
| 107 |
```
|
| 108 |
|
| 109 |
-
## Dataset Statistics
|
| 110 |
For NoMIRACL dataset statistics, please refer to our publication [here](https://arxiv.org/abs/2312.11361).
|
| 111 |
|
| 112 |
|
|
|
|
| 18 |
- te
|
| 19 |
- th
|
| 20 |
- zh
|
| 21 |
+
- pt
|
| 22 |
multilinguality:
|
| 23 |
- multilingual
|
| 24 |
pretty_name: NoMIRACL
|
|
|
|
| 32 |
- apache-2.0
|
| 33 |
---
|
| 34 |
|
| 35 |
+
# Dataset Card for NoMIRACL
|
| 36 |
|
| 37 |
Retrieval Augmented Generation (RAG) is a powerful approach to incorporate external knowledge into large language models (LLMs) to enhance the accuracy and faithfulness of generated responses. However, evaluating LLM robustness in RAG across different language families has been a challenge, leading to gaps in understanding the model's performance against errors in external retrieved knowledge. To address this, we present NoMIRACL, a human-annotated dataset designed for evaluating LLM robustness in RAG across 18 diverse languages.
|
| 38 |
|
|
|
|
| 66 |
Under folders `data/{lang}`,
|
| 67 |
the subset of corpus is saved in `.jsonl.gz` format, with each line to be:
|
| 68 |
```
|
| 69 |
+
{"docid": "28742#27",
|
| 70 |
+
"title": "Supercontinent",
|
| 71 |
"text": "Oxygen levels of the Archaean Eon were negligible and today they are roughly 21 percent. [ ... ]"}
|
| 72 |
```
|
| 73 |
|
|
|
|
| 100 |
query = data['query']
|
| 101 |
positive_passages = data['positive_passages']
|
| 102 |
negative_passages = data['negative_passages']
|
| 103 |
+
|
| 104 |
for entry in positive_passages: # OR 'negative_passages'
|
| 105 |
docid = entry['docid']
|
| 106 |
title = entry['title']
|
| 107 |
text = entry['text']
|
| 108 |
```
|
| 109 |
|
| 110 |
+
## Dataset Statistics
|
| 111 |
For NoMIRACL dataset statistics, please refer to our publication [here](https://arxiv.org/abs/2312.11361).
|
| 112 |
|
| 113 |
|