Computes the cosine similarity matrix between two 1D or 2D tensors.

Returns a matrix of shape {num_a, num_b} where result[i, j] is the cosine similarity between a[i] and b[j].

Computes the dot product similarity matrix between two 1D or 2D tensors.

Returns a matrix of shape {num_a, num_b} where result[i, j] is the dot product between a[i] and b[j].

Computes the Euclidean similarity matrix between two tensors. Returns negative Euclidean distance -dist(a[i], b[j]) so higher values indicate closer vectors.

Computes the Manhattan (L1) similarity matrix between two tensors. Returns negative Manhattan distance -dist(a[i], b[j]) so higher values indicate closer vectors.

Computes the absolute normalized angle distance between pairs of tensors x[i] and y[i].

Matches sentence_transformers.util.pairwise_angle_sim.

Computes the pairwise cosine similarity cos_sim(a[i], b[i]).

Computes the pairwise dot product dot_score(a[i], b[i]).

Computes the pairwise Euclidean similarity -dist(a[i], b[i]).

Computes the pairwise Manhattan similarity -dist(a[i], b[i]).