Commit 路
bdf3ec5
1
Parent(s): 3114e01
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,14 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
Trained on the SQuAD v1.1 dataset from the MRQA Shared Task. The public dev set was divided into two: one for dev and one for test.
|
| 2 |
|
| 3 |
-
Dev results:
|
| 4 |
"eval_exact_match": 88.15914715400723,
|
| 5 |
"eval_f1": 93.91715796563734,
|
| 6 |
"eval_samples": 5291
|
| 7 |
|
| 8 |
-
Test results:
|
| 9 |
"test_exact_match": 86.52455272173582,
|
| 10 |
"test_f1": 92.92134442432088
|
| 11 |
"predict_samples": 5294
|
| 12 |
|
| 13 |
-
More info in the paper:
|
|
|
|
| 14 |
https://arxiv.org/abs/2112.01922
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
+
tags:
|
| 5 |
+
- question-answering
|
| 6 |
+
- qa
|
| 7 |
+
license: "apache-2.0"
|
| 8 |
+
datasets:
|
| 9 |
+
- squad
|
| 10 |
+
metrics:
|
| 11 |
+
- squad
|
| 12 |
+
---
|
| 13 |
+
# Description
|
| 14 |
Trained on the SQuAD v1.1 dataset from the MRQA Shared Task. The public dev set was divided into two: one for dev and one for test.
|
| 15 |
|
| 16 |
+
# Dev results:
|
| 17 |
"eval_exact_match": 88.15914715400723,
|
| 18 |
"eval_f1": 93.91715796563734,
|
| 19 |
"eval_samples": 5291
|
| 20 |
|
| 21 |
+
# Test results:
|
| 22 |
"test_exact_match": 86.52455272173582,
|
| 23 |
"test_f1": 92.92134442432088
|
| 24 |
"predict_samples": 5294
|
| 25 |
|
| 26 |
+
More info in the paper:
|
| 27 |
+
**MetaQA: Combining Expert Agents for Multi-Skill Question Answering**
|
| 28 |
https://arxiv.org/abs/2112.01922
|