fact.yaml

fact:
  prompt: |-
    You are comparing a submitted answer to an expert answer on a given question. Here is the data:
    [BEGIN DATA]
    ************
    [Question]: {input}
    ************
    [Expert]: {ideal}
    ************
    [Submission]: {completion}
    ************
    [END DATA]

    Compare the factual content of the submitted answer with the expert answer. Ignore any differences in style, grammar, or punctuation.
    The submitted answer may either be a subset or superset of the expert answer, or it may conflict with it. Determine which case applies. Answer the question by selecting one of the following options:
    (A) The submitted answer is a subset of the expert answer and is fully consistent with it.
    (B) The submitted answer is a superset of the expert answer and is fully consistent with it.
    (C) The submitted answer contains all the same details as the expert answer.
    (D) There is a disagreement between the submitted answer and the expert answer.
    (E) The answers differ, but these differences don't matter from the perspective of factuality.
  choice_strings: ABCDE
  input_outputs:
    input: completion