NotoriousH2/gsm8k-ko
收藏Hugging Face2026-04-26 更新2026-05-03 收录
下载链接：
https://hf-mirror.com/datasets/NotoriousH2/gsm8k-ko
下载链接
链接失效反馈官方服务：
资源简介：
---
license: mit
language:
- ko
task_categories:
- question-answering
- text-generation
tags:
- gsm8k
- math
- korean
configs:
- config_name: full
data_files:
- split: train
path: full/train.jsonl
- split: test
path: full/test.jsonl
- config_name: sft
data_files:
- split: train
path: sft/train.jsonl
- config_name: eval
data_files:
- split: test
path: eval/test.jsonl
---
# GSM8K Korean
Korean GSM8K-style dataset generated from GSM8K. Failed rows from the source generation outputs were excluded.
## Configs
- `full`: full translation/generation metadata for ok rows only.
- `sft`: train split with `question`, `answer`, `parsed_answer`, and original GSM8K fields.
- `eval`: test split with `question`, short `answer`, and original GSM8K fields.
## Row counts
- full train: 7,309
- full test: 1,296
- sft train: 7,309
- eval test: 1,296
Excluded failed rows: 1,184 train and 187 test.
提供机构：
NotoriousH2
