Commit
•
af152ee
1
Parent(s):
1bc45b3
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
This model is ANCE-Tele trained on MS MARCO. The training details and evaluation results are as follows:
|
| 6 |
+
|
| 7 |
+
|Model|Pretrain Model|Train w/ Marco Title|Marco Dev MRR@10|BEIR Avg NDCG@10|
|
| 8 |
+
|:----|:----|:----|:----|:----|
|
| 9 |
+
|ANCE-Tele|[cocodr-base](https://huggingface.co/OpenMatch/cocodr-base)|w/o|37.3|44.2|
|
| 10 |
+
|
| 11 |
+
|BERI Dataset|NDCG@10|
|
| 12 |
+
|:----|:----|
|
| 13 |
+
|TREC-COVID|77.4|
|
| 14 |
+
|NFCorpus|34.4 |
|
| 15 |
+
|FiQA|29.0 |
|
| 16 |
+
|ArguAna|45.6 |
|
| 17 |
+
|Touché-2020|22.3 |
|
| 18 |
+
|Quora|85.8 |
|
| 19 |
+
|SCIDOCS|14.6 |
|
| 20 |
+
|SciFact|71.0 |
|
| 21 |
+
|NQ|50.5 |
|
| 22 |
+
|HotpotQA|58.8 |
|
| 23 |
+
|Signal-1M|27.2 |
|
| 24 |
+
|TREC-NEWS|34.7 |
|
| 25 |
+
|DBPedia-entity|36.2 |
|
| 26 |
+
|Fever|71.4 |
|
| 27 |
+
|Climate-Fever|17.9 |
|
| 28 |
+
|BioASQ|42.1 |
|
| 29 |
+
|Robust04|41.4 |
|
| 30 |
+
|CQADupStack|34.9 |
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
The implementation is the same as our EMNLP 2022 paper ["Reduce Catastrophic Forgetting of Dense Retrieval Training with Teleportation Negatives"](https://arxiv.org/pdf/2210.17167.pdf). The associated GitHub repository is available at https://github.com/OpenMatch/ANCE-Tele.
|
| 34 |
+
|
| 35 |
+
```
|
| 36 |
+
@inproceedings{sun2022ancetele,
|
| 37 |
+
title={Reduce Catastrophic Forgetting of Dense Retrieval Training with Teleportation Negatives},
|
| 38 |
+
author={Si, Sun and Chenyan, Xiong and Yue, Yu and Arnold, Overwijk and Zhiyuan, Liu and Jie, Bao},
|
| 39 |
+
booktitle={Proceedings of EMNLP 2022},
|
| 40 |
+
year={2022}
|
| 41 |
+
}
|
| 42 |
+
```
|