Is Cosine-Similarity of Embeddings Really About Similarity? · HackerPapers discussions

Is Cosine-Similarity of Embeddings Really About Similarity? — mentioned by ghm2180

Thu, 30 Oct 2025 13:37:52 GMT

There is actually a good analytical result on how vector similarity can easily fail to recover relevant information https://arxiv.org/pdf/2403.05440 > For some linear models the similarities are not even unique, while for others they are implicitly controlled by the regularization. I am not strong in mathematics but if this paper claims run opposite to each other.

Paper and discussion history. A mention is not an endorsement.

Is Cosine-Similarity of Embeddings Really About Similarity? — mentioned by thesehands

Tue, 08 Oct 2024 18:37:31 GMT

This paper is always useful to remember when someone tells you to just use the cosine similarity: https://arxiv.org/abs/2403.05440 Sure, it’s useful but make sure it’s appropriate for your embeddings and remember to run evals to check that things are ‘similar’ in the way that you want them to appear similar. Eg is red similar to green just because they are colors

Paper and discussion history. A mention is not an endorsement.

Is Cosine-Similarity of Embeddings Really About Similarity? — HN submission

Tue, 12 Mar 2024 02:29:22 GMT

Is Cosine-Similarity of Embeddings Really About Similarity?

210 points; 115 discussion replies on HN. Paper and source history. A submission is not an endorsement.