def reasoning_steps_reward()

in src/open_r1/rewards.py [0:0]

• 6 lines of code
• 5 McCabe index (conditional complexity)

def reasoning_steps_reward(completions, **kwargs):
    r"""Reward function that checks for clear step-by-step reasoning.
    Regex pattern:
        Step \d+: - matches "Step 1:", "Step 2:", etc.
        ^\d+\. - matches numbered lists like "1.", "2.", etc. at start of line
        \n- - matches bullet points with hyphens
        \n\* - matches bullet points with asterisks
        First,|Second,|Next,|Finally, - matches transition words
    """
    pattern = r"(Step \d+:|^\d+\.|\n-|\n\*|First,|Second,|Next,|Finally,)"
    completion_contents = [completion[0]["content"] for completion in completions]
    matches = [len(re.findall(pattern, content)) for content in completion_contents]

    # Magic number 3 to encourage 3 steps and more, otherwise partial reward
    return [min(1.0, count / 3) for count in matches]