File size: 1,183 Bytes
9c5e5a3 349a655 9c5e5a3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 | ---
license: mit
---
This model is DPR trained on MS MARCO. The training details and evaluation results are as follows:
|Model|Pretrain Model|Train w/ Marco Title|Marco Dev MRR@10|BEIR Avg NDCG@10|
|:----|:----|:----|:----|:----|
|DPR|bert-base-uncased|w/|32.4|35.5|
|BERI Dataset|NDCG@10|
|:----|:----|
|TREC-COVID|58.8|
|NFCorpus|23.4|
|FiQA|20.6|
|ArguAna|39.4|
|Touché-2020|22.3|
|Quora|78.0|
|SCIDOCS|11.9|
|SciFact|49.4|
|NQ|43.9|
|HotpotQA|45.3|
|Signal-1M|20.2|
|TREC-NEWS|31.8|
|DBPedia-entity|28.7|
|Fever|65.0|
|Climate-Fever|14.9|
|BioASQ|24.1|
|Robust04|32.3|
|CQADupStack|28.3|
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.
```
@inproceedings{sun2022ancetele,
title={Reduce Catastrophic Forgetting of Dense Retrieval Training with Teleportation Negatives},
author={Si, Sun and Chenyan, Xiong and Yue, Yu and Arnold, Overwijk and Zhiyuan, Liu and Jie, Bao},
booktitle={Proceedings of EMNLP 2022},
year={2022}
}
``` |