Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
By specifying a range for similarity scores when mining hard negatives, this argument ensures that negative examples fall within a desired difficulty level. This fine-tuned control helps in avoiding extremes—negatives that are either too close or too far in meaning from the query.
This is also explained in the paper (https://arxiv.org/pdf/2405.05374 (Appendix Algorithm 1: Tunable Negative Mining)), and I also used this code to mine hard negatives, and as a result, I was able to create a Reranker model that performed better than the hard negatives mined with the existing code. I would like to contribute to others using this code to create good models.