tintnguyen's picture
Create README.md
25ade5d verified
---
language:
- vi
size_categories:
- 10M<n<100M
---
This is a filtered subset of the `triples.train.ids.small.tsv` file from the [unicamp-dl/mmarco](https://huggingface.co/datasets/unicamp-dl/mmarco) dataset.
## Selection
Filtered using a custom-trained crossencoder.
Each query-positive pair in the original `triples.train.ids.small.tsv` was scored, and only pairs with a score greater than **0.5** were retained.