A thin, shape-aware wrapper around LLMR::llm_agreement(). Pass either a
gold_set() whose coders columns hold individual human codings (the
human-human reliability a methods section must report), or any data frame
plus the label columns to compare (e.g. model replicates from
code_corpus(), or one human column and one model column).
Arguments
- x
A
gold_set()or a data frame.- cols
For data frames: character vector of label columns. Ignored for gold sets (their
coderscolumns are used).
Value
An llmr_agreement object (mean pairwise agreement,
Krippendorff's alpha, per-unit majorities); see
LLMR::llm_agreement().
