OpinionQA alignment per row: 1 - W1 / (N - 1) on the ordinal scale (Santurkar et al., 2023)

OpinionQA alignment per row: 1 - W1 / (N - 1) on the ordinal scale (Santurkar et al., 2023)

Usage

align_opinionqa_alignment(human, model, positions = NULL)

Arguments

human, model

Distributions: cells, a matrix with one row per distribution, or a vector.

positions

Option positions, or NULL for 0, 1, 2, ...

Value

A numeric vector.