One row per unit per cell: which label each text received under each
specification. response_id is the join key into an archive (see the
archive workflow) and is NA when the runner did not report one. Cell summaries answer "is the
estimate stable"; this table answers "which units moved when it was
not" – and without it an audit is unfalsifiable.
Arguments
- audit
An
audit_run()result.
Examples
plan <- audit_plan(data.frame(text = c("a", "b")), "text",
function(d) mean(d$label == "yes"), c("yes", "no"), "{text}")
plan <- audit_add_models(plan,
list(demo = LLMR::llm_config("groq", "demo", temperature = 0)))
runner <- function(x, ...) transform(x, response_text = c("yes", "no"))
audit <- audit_run(plan, .runner = runner)
audit_units(audit)