Skip to contents

The inputs of every rated call of fn and the right answer (the model's, when it was rated right; the correction, when wrong), typed like the function's inputs and outputs: ready for evaluate() and the optimizers. Calls whose data has another shape (the inputs or outputs changed since) are left out and counted (other_signature); calls that record the same data pool, even when the instruction, a default or reasoning changed. So are calls whose inputs the log did not keep (no_content, log_content), and rated calls that give no answer (no_answer: a right verdict on an answer that was not kept). Reads every language's log, formats 1 and 2.

Usage

rated(fn, by = NULL, folder = NULL, since = NULL, any_file = FALSE)

Arguments

fn

An AI function, or a program.

by

Only this person's ratings.

folder

The log folder.

since

Only calls and ratings from this time on.

any_file

A function defined in a notebook or a script is known by its file too, so two notebooks' summarize are two programs. TRUE takes its calls from any file (a notebook that moved).

Value

A tibble: the inputs, the outputs, then call, version, rating, rated_by, origin, sample, disputed. Attribute left_out: the counts.