Talk:Hard negative mining

Latest comment: 13 days ago by Cerlonghish in topic Reviewer 2

Suggestions

edit

Reviewer 1

edit

- Corrected sentence paragraph, as not only there must be many negatives but many easy negatives.

- Added bootstrapping current definition to avoid confusion.

- Added that negatives need to be far in latent space from positives (just a clarification).

- Tried to add first image caption since it was too short and not clear.

- Added some "for instance" if an example is used to explain the topic since the topic might be broader than just the example.

- Few minor text edits for better grammar.

I would suggest to define the terms in the contrastive self-supervised learning formula.

Overall good job! Checco Raggio (talk) 13:14, 3 July 2026 (UTC)Reply

Mathematical notation in this article is definitely needed to be added and better defined.
Thank you for the edits and the review! Cerlonghish (talk) 13:20, 3 July 2026 (UTC)Reply

Reviewer 3

edit

thank you! I think that almost all the work has been made by reviewer 1. very good article btw Dani1998ele (talk) 13:31, 3 July 2026 (UTC)Reply

Thank you! Cerlonghish (talk) 14:25, 3 July 2026 (UTC)Reply

Reviewer 2

edit

As already noted by the other two reviewers, this is a great piece of work, and Reviewer 1 has already addressed most of the substantive points. I only had a minor clarification regarding the term sampling bias, which I have already fixed in the text. ThetaXVIII (talk) 14:01, 3 July 2026 (UTC)Reply

Thank you! Cerlonghish (talk) 14:25, 3 July 2026 (UTC)Reply