evalstone/evalscope/docs/zh/benchmarks/drivel_writing.md

DrivelologyNarrativeWriting

概述

Drivelology 叙事写作评估模型生成详细描述的能力，以阐释“drivelology”文本中隐含的叙事——这类语言表达在句法上连贯，但在语用上却具有悖论性、情感负载性或修辞颠覆性。

任务描述

• 任务类型：叙事生成与评估
• 输入：Drivelology 文本样本
• 输出：生成的叙事描述，解释文本的隐含意义
• 领域：语言学分析、叙事生成

核心特点

• 测试模型生成叙事解释的能力
• 要求理解多层次的语言含义
• 使用 LLM-as-judge 方法与参考叙事进行对比评估
• 采用李克特量表（1-5 分）对匹配质量评分
• 考察模型在语言和文化理解方面的深度

评估说明

• 默认配置使用 0-shot 评估
• 采用 LLM-as-judge 进行评估
• 指标：平均李克特得分（1-5 分制）
• 评估生成叙事的相关性、准确性、深度和细节

属性

属性	值
基准测试名称	drivel_writing
数据集ID	extraordinarylab/drivel-hub
论文	N/A
标签	Knowledge, Reasoning
指标	bert_score, gpt_score
默认示例数	0-shot
评估划分	test

数据统计

指标	值
总样本数	600
提示词长度（平均）	313.18 字符
提示词长度（最小/最大）	256 / 717 字符

样例示例

子集: narrative-writing-english

{
  "input": [
    {
      "id": "f47953a9",
      "content": [
        {
          "text": "You need to first read and understand the text given. Generate a detailed description to illustrate the implicit narrative of the text.\n\nPlease provide your response in English, with a clear and comprehensive explanation of the narrative.\n\nText: 後天的努力比什麼都重要，所以今天和明天休息。"
        }
      ]
    }
  ],
  "target": "This creates a paradoxical tone, as it acknowledges the value of diligence but simultaneously advocates for procrastination. The underlying message could reflect a lighthearted take on balancing work and rest or even poking fun at the tendency to delay responsibilities.",
  "id": 0,
  "group_id": 0,
  "metadata": {
    "text": "後天的努力比什麼都重要，所以今天和明天休息。",
    "reference_narrative": "This creates a paradoxical tone, as it acknowledges the value of diligence but simultaneously advocates for procrastination. The underlying message could reflect a lighthearted take on balancing work and rest or even poking fun at the tendency to delay responsibilities."
  }
}

提示模板

提示模板：

You need to first read and understand the text given. Generate a detailed description to illustrate the implicit narrative of the text.

Please provide your response in English, with a clear and comprehensive explanation of the narrative.

Text: {text}

使用方法

使用 CLI

evalscope eval \
    --model YOUR_MODEL \
    --api-url OPENAI_API_COMPAT_URL \
    --api-key EMPTY_TOKEN \
    --datasets drivel_writing \
    --limit 10  # 正式评估时请删除此行

使用 Python

from evalscope import run_task
from evalscope.config import TaskConfig

task_cfg = TaskConfig(
    model='YOUR_MODEL',
    api_url='OPENAI_API_COMPAT_URL',
    api_key='EMPTY_TOKEN',
    datasets=['drivel_writing'],
    limit=10,  # 正式评估时请删除此行
)

run_task(task_cfg=task_cfg)