115 lines
2.8 KiB
Markdown
115 lines
2.8 KiB
Markdown
# Winogrande
|
||
|
||
|
||
## 概述
|
||
|
||
Winogrande 是一个大规模的常识推理基准测试，专门用于在 Winograd Schema Challenge 格式下测试代词消歧任务。它包含 4.4 万个问题，要求模型理解物理和社交方面的常识。
|
||
|
||
## 任务描述
|
||
|
||
- **任务类型**：代词消歧 / 常识推理
|
||
- **输入**：包含歧义代词的句子及两个选项
|
||
- **输出**：正确选项（A 或 B），用于消解代词指代
|
||
- **格式**：在两个名词短语之间进行二选一
|
||
|
||
## 主要特点
|
||
|
||
- 包含 4.4 万个 Winograd 风格的代词消歧问题
|
||
- 通过对抗性过滤减少数据集偏差
|
||
- 测试物理常识（物体属性、动作等）
|
||
- 测试社交常识（意图、情感等）
|
||
- 需要理解上下文以解决歧义
|
||
|
||
## 评估说明
|
||
|
||
- 默认配置使用 **0-shot** 评估
|
||
- 采用二选一格式（选项1 vs 选项2）
|
||
- 答案转换为 A/B 字母格式
|
||
- 使用简单准确率（accuracy）作为评估指标
|
||
- 常用于常识推理能力评估
|
||
|
||
## 属性
|
||
|
||
| 属性 | 值 |
|
||
|----------|-------|
|
||
| **基准测试名称** | `winogrande` |
|
||
| **数据集ID** | [AI-ModelScope/winogrande_val](https://modelscope.cn/datasets/AI-ModelScope/winogrande_val/summary) |
|
||
| **论文** | N/A |
|
||
| **标签** | `MCQ`, `Reasoning` |
|
||
| **指标** | `acc` |
|
||
| **默认示例数** | 0-shot |
|
||
| **评估分割** | `validation` |
|
||
|
||
## 数据统计
|
||
|
||
| 指标 | 值 |
|
||
|--------|-------|
|
||
| 总样本数 | 1,267 |
|
||
| 提示词长度（平均） | 306.58 字符 |
|
||
| 提示词长度（最小/最大） | 271 / 384 字符 |
|
||
|
||
## 样例示例
|
||
|
||
**子集**: `default`
|
||
|
||
```json
|
||
{
|
||
"input": [
|
||
{
|
||
"id": "f64dd7e2",
|
||
"content": "Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: [LETTER]' (without quotes) where [LETTER] is one of A,B.\n\nSarah was a much better surgeon than Maria so _ always got the easier cases.\n\nA) Sarah\nB) Maria"
|
||
}
|
||
],
|
||
"choices": [
|
||
"Sarah",
|
||
"Maria"
|
||
],
|
||
"target": "B",
|
||
"id": 0,
|
||
"group_id": 0,
|
||
"metadata": {
|
||
"id": "winogrande_xs val 0"
|
||
}
|
||
}
|
||
```
|
||
|
||
## 提示模板
|
||
|
||
**提示模板:**
|
||
```text
|
||
Answer the following multiple choice question. The entire content of your response should be of the following format: 'ANSWER: [LETTER]' (without quotes) where [LETTER] is one of {letters}.
|
||
|
||
{question}
|
||
|
||
{choices}
|
||
```
|
||
|
||
## 使用方法
|
||
|
||
### 使用 CLI
|
||
|
||
```bash
|
||
evalscope eval \
|
||
--model YOUR_MODEL \
|
||
--api-url OPENAI_API_COMPAT_URL \
|
||
--api-key EMPTY_TOKEN \
|
||
--datasets winogrande \
|
||
--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=['winogrande'],
|
||
limit=10, # 正式评估时请删除此行
|
||
)
|
||
|
||
run_task(task_cfg=task_cfg)
|
||
``` |