evalstone/evalscope/docs/zh/benchmarks/bhashabenchv1_legal.md
sora 13274243a0 Bump vendored EvalScope and add K3-ready DPV4 configs.
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>
2026-09-02 07:30:48 +00:00

129 lines
3.8 KiB
Markdown
Raw Permalink Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# BhashaBench-V1 (Legal)
## 概述
BhashaBench-Legal 是 BhashaBench-Multi 法律领域的前身:这是一个领域特定的多项选择题基准测试,用于评估大语言模型对印度法律的知识,涵盖英语和印地语。
## 任务描述
- **任务类型**:领域特定的多项选择题问答
- **输入**一道包含4个选项的印度法律问题语言为英语或印地语
- **输出**:正确答案的字母
- **语言**:英语、印地语
## 主要特点
- 每种语言包含5,600至17,000道题目仅涵盖英语和印地语
- 作为 BhashaBench-Multi 的前身:领域相同,但语言覆盖范围更窄
- 每个领域对应一个独立的代码仓库,英语和印地语分别作为独立的配置项
## 评估说明
- 默认配置使用 **0-shot** 评估(仅提供 test 分割)
- 使用 `subset_list` 可评估单一语言(例如 `['Hindi']`
- 需要访问此受限制的数据集——在 ModelScope默认 Hub请先接受条款并确保已登录或者将 `dataset_hub` 设置为 `huggingface`,并在 huggingface.co 上接受条款后使用 `HF_TOKEN`
- 如需同一领域但更广泛的语言覆盖,请参见 `bhasha_bench_multi_legal`涵盖22种印度语言无需授权
## 属性
| 属性 | 值 |
|----------|-------|
| **基准测试名称** | `bhashabenchv1_legal` |
| **数据集ID** | [bharatgenai/BhashaBench-Legal](https://modelscope.cn/datasets/bharatgenai/BhashaBench-Legal/summary) |
| **论文** | 无 |
| **标签** | `Knowledge`, `MCQ`, `MultiLingual` |
| **指标** | `accuracy` |
| **默认示例数** | 0-shot |
| **评估分割** | `test` |
## 数据统计
| 指标 | 值 |
|--------|-------|
| 总样本数 | 24,365 |
| 提示词长度(平均) | 513.88 字符 |
| 提示词长度(最小/最大) | 229 / 4628 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 |
|--------|---------|-------------|------------|------------|
| `English` | 17,047 | 539.36 | 233 | 4628 |
| `Hindi` | 7,318 | 454.52 | 229 | 1748 |
## 样例示例
**子集**: `English`
```json
{
"input": [
{
"id": "6e1ae42b",
"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\nPower to amend the issue or frame additional issues prior to passing of a decree vests in a Court by virtue of which provision of the Code of Civil Procedure, 1908?\n\nA) Order XIV Rule 1\nB) Order XIV Rule 5\nC) Order XIV Rule 6\nD) Section 151"
}
],
"choices": [
"Order XIV Rule 1",
"Order XIV Rule 5",
"Order XIV Rule 6",
"Section 151"
],
"target": "B",
"id": 0,
"group_id": 0,
"metadata": {
"language": "English",
"topic": "Procedural Law"
}
}
```
## 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 bhashabenchv1_legal \
--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=['bhashabenchv1_legal'],
dataset_args={
'bhashabenchv1_legal': {
# subset_list: ['English', 'Hindi'] # 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)
```