Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,31 @@
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
4 |
+
|
5 |
+
## EQA-PMR-large
|
6 |
+
EQA-PMR-large is initialized with [PMR-large](https://huggingface.co/DAMO-NLP-SG/PMR-large) and further fine-tuned on 6 Extractive Question Answering (EQA) training data from training split of [MRQA](https://aclanthology.org/D19-5801).
|
7 |
+
|
8 |
+
The model performance on the in-dev sets are:
|
9 |
+
|
10 |
+
|| SQuAD | NewsQA | HotpotQA | NaturalQuestions | TriviaQA | SearchQA|
|
11 |
+
|--|------------|-----------|----------|--|
|
12 |
+
|RoBERTa-large (single-task model)| 94.2 | 73.8 | 81.6|83.3| 85.1 | 85.7 |
|
13 |
+
|PMR-large (single-task model)| 94.5 | 74.0 | 83.6 | 83.8 | 85.1 | 88.3 |
|
14 |
+
|EQA-PMR-large (multi-task model)| 94.2 | 73.7 | 66.9 | 82.3 | 85.4 | 88.7 |
|
15 |
+
|
16 |
+
Note that the performance of RoBERTa-large and PMR-large are single-task fine-tuning, while EQA-PMR-large is a multi-task fine-tuned model.
|
17 |
+
As it is fine-tuned on multiple datasets, we believe that EQA-PMR-large has a better generalization capability to other EQA tasks than PMR-large and RoBERTa-large.
|
18 |
+
|
19 |
+
### How to use
|
20 |
+
You can try the codes from [this repo](https://github.com/DAMO-NLP-SG/PMR/QA) for both training and inference.
|
21 |
+
|
22 |
+
|
23 |
+
### BibTeX entry and citation info
|
24 |
+
```bibtxt
|
25 |
+
@article{xu2022clozing,
|
26 |
+
title={From Clozing to Comprehending: Retrofitting Pre-trained Language Model to Pre-trained Machine Reader},
|
27 |
+
author={Xu, Weiwen and Li, Xin and Zhang, Wenxuan and Zhou, Meng and Bing, Lidong and Lam, Wai and Si, Luo},
|
28 |
+
journal={arXiv preprint arXiv:2212.04755},
|
29 |
+
year={2022}
|
30 |
+
}
|
31 |
+
```
|