130 lines
3.6 KiB
Markdown
130 lines
3.6 KiB
Markdown
# DocMath
|
|
|
|
## 概述
|
|
|
|
DocMath-Eval 是一个专注于特定领域内数值推理的综合性基准测试。该基准要求模型理解长篇且专业的文档，并基于文档内容进行数值推理以回答问题。
|
|
|
|
## 任务描述
|
|
|
|
- **任务类型**：基于文档的数学推理
|
|
- **输入**：长文档上下文 + 数值推理问题
|
|
- **输出**：带推理过程的数值答案
|
|
- **重点**：长上下文理解与定量推理能力
|
|
|
|
## 主要特点
|
|
|
|
- 需要理解长篇专业文档
|
|
- 在文档上下文中进行数值推理
|
|
- 多种复杂度级别（复杂/简单，长/短）
|
|
- 测试真实场景下的文档理解能力
|
|
- 同时考察阅读理解与数学技能
|
|
|
|
## 评估说明
|
|
|
|
- 默认配置使用 **0-shot** 评估
|
|
- 使用 LLM-as-judge 进行答案评估
|
|
- 子集包括：`complong_testmini`、`compshort_testmini`、`simplong_testmini`、`simpshort_testmini`
|
|
- 答案格式：`"Therefore, the answer is (answer)"`
|
|
|
|
## 属性
|
|
|
|
| 属性 | 值 |
|
|
|----------|-------|
|
|
| **基准测试名称** | `docmath` |
|
|
| **数据集ID** | [yale-nlp/DocMath-Eval](https://modelscope.cn/datasets/yale-nlp/DocMath-Eval/summary) |
|
|
| **论文** | N/A |
|
|
| **标签** | `LongContext`, `Math`, `Reasoning` |
|
|
| **指标** | `acc` |
|
|
| **默认示例数** | 0-shot |
|
|
| **评估分割** | `test` |
|
|
|
|
## 数据统计
|
|
|
|
| 指标 | 值 |
|
|
|--------|-------|
|
|
| 总样本数 | 800 |
|
|
| 提示词长度（平均） | 68791.03 字符 |
|
|
| 提示词长度（最小/最大） | 505 / 1009038 字符 |
|
|
|
|
**各子集统计数据：**
|
|
|
|
| 子集 | 样本数 | 提示平均长度 | 提示最小长度 | 提示最大长度 |
|
|
|--------|---------|-------------|------------|------------|
|
|
| `complong_testmini` | 300 | 175355.17 | 18687 | 1009038 |
|
|
| `compshort_testmini` | 200 | 1990.74 | 505 | 9460 |
|
|
| `simplong_testmini` | 100 | 13972.84 | 6870 | 24001 |
|
|
| `simpshort_testmini` | 200 | 3154.2 | 560 | 9600 |
|
|
|
|
## 样例示例
|
|
|
|
**子集**: `complong_testmini`
|
|
|
|
```json
|
|
{
|
|
"input": [
|
|
{
|
|
"id": "a07cbfcf",
|
|
"content": "Please read the following text and answer the question below.\n\n<text>\nDELTA AIR LINES, INC.\nConsolidated Balance Sheets\n| (in millions, except share data) | March 31, 2018 | December 31, 2017 |\n| ASSETS |\n| Current Assets: |\n| Cash and cash e ... [TRUNCATED] ... comprehensive income for foreign currency exchange contracts in 2017 and 2018, and the changes in value for derivative contracts and other in 2018, in million?\n\nFormat your response as follows: \"Therefore, the answer is (insert answer here)\"."
|
|
}
|
|
],
|
|
"target": "-31.0",
|
|
"id": 0,
|
|
"group_id": 0,
|
|
"metadata": {
|
|
"question_id": "complong-testmini-0",
|
|
"answer_type": "float"
|
|
}
|
|
}
|
|
```
|
|
|
|
*注：部分内容因展示需要已被截断。*
|
|
|
|
## 提示模板
|
|
|
|
**提示模板：**
|
|
```text
|
|
Please read the following text and answer the question below.
|
|
|
|
<text>
|
|
{context}
|
|
</text>
|
|
|
|
{question}
|
|
|
|
Format your response as follows: "Therefore, the answer is (insert answer here)".
|
|
```
|
|
|
|
## 使用方法
|
|
|
|
### 使用 CLI
|
|
|
|
```bash
|
|
evalscope eval \
|
|
--model YOUR_MODEL \
|
|
--api-url OPENAI_API_COMPAT_URL \
|
|
--api-key EMPTY_TOKEN \
|
|
--datasets docmath \
|
|
--limit 10 # 正式评估时请删除此行
|
|
```
|
|
|
|
### 使用 Python
|
|
|
|
```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=['docmath'],
|
|
dataset_args={
|
|
'docmath': {
|
|
# subset_list: ['complong_testmini', 'compshort_testmini', 'simplong_testmini'] # 可选，用于评估特定子集
|
|
}
|
|
},
|
|
limit=10, # 正式评估时请删除此行
|
|
)
|
|
|
|
run_task(task_cfg=task_cfg)
|
|
``` |