Hi, I'm working on a project where I am reranking text matches with vector search. That means that I use a query similar to this:

{
  "q": "shoes",
  "query_by": "title,brand",
  "sort_by": "_text_match:desc,_vector_query(embedding:([0.43, 0.13, 0.21])):asc"
}

I'm using v27.rc35. It shows the vector distances. Can anybody explain how the sorting and the scores work? It seems well documented for cosine similarity, but I'm using inner product, and not much is explained. With cosine similarity values range from 0-2, where 1=0 neutral and 0=similar. With inner product, I get values like -27. How am I to interpret these values?