Is the L2 distance score in vector search an inverted score?

[yy0vnne]

Hi there, the L2 scores and cosine scores for LanceDB vector search suggests that smaller _distance scores means more similar.
But the general understanding is that for L2 scores, smaller _distance means more similar, while for cosine score, greater _distance means more similar. May I check if L2 distance is an inverted score in LanceDB?

[westonpace]

The L2 score is not inverted.

The cosine score is "cosine distance" and not "cosine similarity".

Cosine distance is defined as 1 - cosine_similarity (cosine distance is inverted cosine similarity).

In all cases a smaller _distance means more similar.