SYNTHETIC-1-SFT-Data-Code_decontaminated
收藏魔搭社区2025-11-27 更新2025-03-01 收录
下载链接:
https://modelscope.cn/datasets/open-r1/SYNTHETIC-1-SFT-Data-Code_decontaminated
下载链接
链接失效反馈官方服务:
资源简介:
## Dataset description
This dataset is the same as [open-r1/SYNTHETIC-1-SFT-Data-Code](https://huggingface.co/datasets/open-r1/SYNTHETIC-1-SFT-Data-Code) decontaminated against the benchmark datasets.
The decontamination has been run using the script in [huggingface/open-r1](https://github.com/huggingface/open-r1/pull/416):
```shell
python scripts/decontaminate.py \
--dataset "open-r1/SYNTHETIC-1-SFT-Data-Code" \
-c
...
Removed 5 samples from 'aime_2025'
Removed 50 samples from 'math_500'
Removed 13234 samples from 'lcb'
Initial size: 62953, Final size: 49664
```
## 数据集描述
本数据集与[open-r1/SYNTHETIC-1-SFT-Data-Code](https://huggingface.co/datasets/open-r1/SYNTHETIC-1-SFT-Data-Code)完全一致,但已针对各类基准数据集完成去污染(decontamination)处理。本次去污染操作通过[huggingface/open-r1](https://github.com/huggingface/open-r1/pull/416)提供的脚本执行:
shell
python scripts/decontaminate.py
--dataset "open-r1/SYNTHETIC-1-SFT-Data-Code"
-c
...
已从`aime_2025`中移除5条样本
已从`math_500`中移除50条样本
已从`lcb`中移除13234条样本
初始样本总量:62953,最终样本总量:49664
提供机构:
maas
创建时间:
2025-02-25



