如何定义代码评估器

def perform_eval(run, example):
    # 访问数据
    inputs = run['inputs']
    outputs = run['outputs']
    reference_outputs = example['outputs']  # 可选：参考/预期输出

    # 在此处编写评估逻辑
    score = ...

    # 返回包含指标名称的字典
    return {"metric_name": score}

def perform_eval(run, example):
    """检查答案是否与预期答案完全匹配。"""
    actual = run['outputs']['answer']
    expected = example['outputs']['answer']

    is_correct = actual == expected
    return {"exact_match": is_correct}

def perform_eval(run, example):
    """检查输入文本是否包含有害语言。"""
    text = run['inputs'].get('text', '').lower()
    toxic_words = ["idiot", "stupid", "hate", "awful"]

    is_toxic = any(word in text for word in toxic_words)
    return {"is_toxic": is_toxic}