13274243a0
Keep K3 suite selection and report-schema scoring in bash, merge K3/vision dataset_args into dpv4 yamls, and pin EvalScope at 735d920ee911 with local patches. Co-authored-by: Cursor <cursoragent@cursor.com>
125 lines
3.7 KiB
Markdown
125 lines
3.7 KiB
Markdown
# Sanskriti
|
||
|
||
|
||
## 概述
|
||
|
||
Sanskriti 是一个多项选择题基准测试，用于评估对印度各邦文化、历史和地理知识的掌握情况。题目基于各邦特有的属性（如艺术、美食、节日等）构建，答案均参考维基百科。该适配器加载的是 SANSKRITI 论文（arXiv:2506.15355）中所用数据集在 ModelScope 上的镜像版本 `evalscope/Sanskriti`。
|
||
|
||
## 任务描述
|
||
|
||
- **任务类型**：多项选择题问答（Multiple-Choice Trivia Question Answering）
|
||
- **输入**：一道关于特定印度邦文化/地理/历史的问题，附带 4 个选项
|
||
- **输出**：正确答案对应的字母（A/B/C/D）
|
||
- **子集**：`association`（邦与属性关联类题目）、`country`（国家层面常识题）、`gk`（通用知识题）、`states`（邦识别类题目）
|
||
|
||
## 评估说明
|
||
|
||
- 默认配置采用 **0-shot** 评估方式（尽管上游数据集中该部分被命名为 `train`，但实际为评估数据）
|
||
- 问题和选项均为英文
|
||
- 论文指出部分题目涉及的文化元素存在模糊性；加载时会跳过少量（约 0.6%）答案字段与四个选项均不匹配的样本
|
||
|
||
## 属性
|
||
|
||
| 属性 | 值 |
|
||
|----------|-------|
|
||
| **基准测试名称** | `sanskriti` |
|
||
| **数据集ID** | [evalscope/Sanskriti](https://modelscope.cn/datasets/evalscope/Sanskriti/summary) |
|
||
| **论文** | N/A |
|
||
| **标签** | `Knowledge`, `MCQ` |
|
||
| **指标** | `accuracy` |
|
||
| **默认示例数** | 0-shot |
|
||
| **评估划分** | `train` |
|
||
|
||
|
||
## 数据统计
|
||
|
||
| 指标 | 值 |
|
||
|--------|-------|
|
||
| 总样本数 | 21,726 |
|
||
| 提示词长度（平均） | 322.93 字符 |
|
||
| 提示词长度（最小/最大） | 256 / 636 字符 |
|
||
|
||
**各子集统计数据：**
|
||
|
||
| 子集 | 样本数 | 提示词平均长度 | 提示词最小长度 | 提示词最大长度 |
|
||
|--------|---------|-------------|------------|------------|
|
||
| `association` | 5,453 | 343.41 | 273 | 523 |
|
||
| `country` | 5,563 | 284.48 | 256 | 417 |
|
||
| `gk` | 5,328 | 346.94 | 263 | 547 |
|
||
| `states` | 5,382 | 318.17 | 260 | 636 |
|
||
|
||
## 样例示例
|
||
|
||
**子集**: `association`
|
||
|
||
```json
|
||
{
|
||
"input": [
|
||
{
|
||
"id": "0629b222",
|
||
"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,C,D.\n\nWhich of the given regions is home to the Jarawa body painting?\n\nA) Surguja district\nB) South Andaman and Middle Andaman Islands\nC) Buddha Marg, Patna\nD) Telangana"
|
||
}
|
||
],
|
||
"choices": [
|
||
"Surguja district",
|
||
"South Andaman and Middle Andaman Islands",
|
||
"Buddha Marg, Patna",
|
||
"Telangana"
|
||
],
|
||
"target": "B",
|
||
"id": 0,
|
||
"group_id": 0,
|
||
"subset_key": "association",
|
||
"metadata": {
|
||
"state": "Andaman_and_Nicobar",
|
||
"attribute": "Art"
|
||
}
|
||
}
|
||
```
|
||
|
||
## 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 sanskriti \
|
||
--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=['sanskriti'],
|
||
dataset_args={
|
||
'sanskriti': {
|
||
# subset_list: ['association', 'country', 'gk'] # 可选，用于指定评估特定子集
|
||
}
|
||
},
|
||
limit=10, # 正式评估时请删除此行
|
||
)
|
||
|
||
run_task(task_cfg=task_cfg)
|
||
```
|