Template

#!/usr/bin/env python
# -*- coding: utf-8 -*-


import json
import re

def check_percent_in_process(data: dict) -> list[bool]:
    """
    判断每个 process 元素是否包含百分号 %
    
    :param data: 包含 'process' 字段的字典
    :return: 每个元素是否包含百分号的布尔列表
    """
    process_list = data.get("process", [])
    return ['%' in step for step in process_list]


def extract_boxed_answer(text: str) -> str:
    # 匹配最后一个 \boxed{...} 中的内容
    matches = re.findall(r'\\boxed\{(.*?)\}', text)
    return matches[-1] if matches else None


def process_jsonl(input_path: str, output_path: str):
    with open(input_path, 'r', encoding='utf-8') as infile, \
         open(output_path, 'w', encoding='utf-8') as outfile:
        
        for line in infile:
            data = json.loads(line)
            distill_text = data.get("distill_answer", "")
            infer_answer = extract_boxed_answer(distill_text).replace('\\%', '%')

            try:
                if check_percent_in_process(data):
                    if infer_answer[-1] == '%':
                        infer_answer = float(infer_answer[:-1])
                    else:
                        infer_answer = float(infer_answer) * 100
                else:
                    if infer_answer[-1] == '%':
                        infer_answer = float(infer_answer[:-1]) * 100
            except:
                infer_answer = None

            if infer_answer:
                data["infer_answer"] = infer_answer

            json.dump(data, outfile, ensure_ascii=False)
            outfile.write('\n')


def main():
    # 示例调用：
    process_jsonl("data/temp.jsonl", "data/a0_infer_answer.jsonl")

if __name__=='__main__':
    main()


#!/usr/bin/env python
# -*- coding: utf-8 -*-
import json

def split_jsonl_by_result(
    input_path: str,
    equal_output_path: str,
    unequal_output_path: str,
    threshold: float = 1e-3
):
    """
    读取 JSONL 文件，比较 result 与 infer_answer 是否相等（相对误差小于阈值），
    并将结果分别保存到两个 JSONL 文件中。

    参数:
    - input_path: 输入 JSONL 文件路径
    - equal_output_path: 相等结果保存路径
    - unequal_output_path: 不相等结果保存路径
    - threshold: 相对误差阈值，默认 1e-3
    """
    equal_count = 0
    unequal_count = 0

    with open(equal_output_path, 'w', encoding='utf-8') as eq_out, \
         open(unequal_output_path, 'w', encoding='utf-8') as neq_out, \
         open(input_path, 'r', encoding='utf-8') as f:

        for line in f:
            try:
                data = json.loads(line)
                result = data.get('result')
                infer_answer = data.get('infer_answer')

                if isinstance(result, (int, float)) and isinstance(infer_answer, (int, float)):
                    relative_error = abs(result - infer_answer) / abs(result)
                    if relative_error < threshold:
                        eq_out.write(json.dumps(data, ensure_ascii=False) + '\n')
                        equal_count += 1
                    else:
                        neq_out.write(json.dumps(data, ensure_ascii=False) + '\n')
                        unequal_count += 1
                else:
                    # 数据格式不正确，默认写入不相等文件
                    neq_out.write(json.dumps(data, ensure_ascii=False) + '\n')
                    unequal_count += 1
            except json.JSONDecodeError:
                print("跳过无法解析的行：", line.strip())
                continue

    print(f"处理完成：相等记录 {equal_count} 条，不相等记录 {unequal_count} 条。")


def main():
    split_jsonl_by_result(
        input_path='data/a0_infer_answer.jsonl',
        equal_output_path='data/a1_equal.jsonl',
        unequal_output_path='data/a1_unequal.jsonl',
        threshold=1e-3
    )


if __name__=='__main__':
    main()