Zaynoid/sft_granularity_cot
收藏Hugging Face2026-04-03 更新2026-04-12 收录
下载链接：
https://hf-mirror.com/datasets/Zaynoid/sft_granularity_cot
下载链接
链接失效反馈官方服务：
资源简介：
---
language:
- en
task_categories:
- token-classification
tags:
- medical
- NER
- chain-of-thought
- clinical
- alpaca
configs:
- config_name: default
data_files:
- split: train
path: data/train.jsonl
---
# sft_granularity_cot
Clinical NER dataset with Chain-of-Thought reasoning for SFT fine-tuning.
| Field | Description |
|---|---|
| `instruction` | System task description |
| `input` | Clinical sentence |
| `output` | `<think>` reasoning block + indented JSON entities |
**Records**: 3943
**Format**: Alpaca (instruction / input / output)
## Output Format
```
<think>
Step-by-step reasoning about each token...
</think>
{
"entities": [
{"text": "WBC", "label": "Test"},
{"text": "9.3", "label": "Test_Result"}
]
}
```
## Critical Rule
`Test` names and `Test_Result` values are **always** separate entities — never combined.
## Load
```python
from datasets import load_dataset
ds = load_dataset("Zaynoid/sft_granularity_cot", split="train")
```
提供机构：
Zaynoid
