Currently, the faiss, numpy and torch backend support binary index/vectors. These backends all use the hamming score as the similarity metric. The hamming scores should be bound between 0 and 1.