INNER CODE UNIT · Python

grade_answers

Future-House/BixBench · grade_outputs.py:49

async def grade_answers(
    input_file: str | Path,
    answer_mode: AnswerMode,
    model_name: str = "gpt-4o",
    temperature: float = 1.0,
    **kwargs: dict[str, Any],
):
    """Grade answers based on evaluation mode."""
    query_df = pd.read_csv(input_file)

    if answer_mode == AnswerMode.openanswer:
        llm_client = LiteLLMModel(
            name=f"{model_name}",
            config={"name": model_name, "temperature": temperature, **kwargs},
        )
        grader = GradeAnswer(
            answer_mode=answer_mode,
            llm_client=llm_client,

View source record →