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. | |