Also known as out-of-sample error, generalisation error
in machine learning, a measure of how accurately an algorithm is able to predict outcome values for previously unseen data
Discovered by embedding cosine similarity (sentence-transformers MiniLM, 384-dim).