JohnTdi/polish-llm-sft-en
收藏Hugging Face2026-05-27 更新2026-05-31 收录
下载链接：
https://hf-mirror.com/datasets/JohnTdi/polish-llm-sft-en
下载链接
链接失效反馈官方服务：
资源简介：
---
license: apache-2.0
language:
- en
task_categories:
- text-generation
- question-answering
- multiple-choice
tags:
- english
- sft
- instruction-tuning
- mcq
- reasoning
- math
- code
size_categories:
- 1K<n<10K
---
# Polish LLM SFT Dataset — English
**EN** | English-language subset of a dataset prepared to help various LLMs learn and improve their Polish language capabilities. Companion to the [Polish repo](https://huggingface.co/datasets/JohnTdi/polish-llm-sft-pl).
**PL** | Anglojęzyczny podzbiór datasetu przygotowanego z myślą o nauczaniu języka polskiego różnych modeli LLM. Companion do [repo polskiego](https://huggingface.co/datasets/JohnTdi/polish-llm-sft-pl).
**4 884 samples** · Apache 2.0 · Language: EN
---
## Format
```json
{"messages": [
{"role": "user", "content": "..."},
{"role": "assistant", "content": "..."}
]}
```
---
## Structure
| Directory | Description | Samples |
|:---|:---|---:|
| `math/` | Math word problems | 267 |
| `mcq_general/` | Multi-domain MCQ | 1 937 |
| `mcq_rc/` | Reading comprehension MCQ | 75 |
| `reasoning/` | Chain-of-thought reasoning | 2 605 |
---
## Quality scale
| Score | Meaning |
|:---:|:---|
| q10 | Handwritten with Claude Opus 4.7, verified |
| q9 | Synthetically generated, ~95% accuracy |
| q8 | Good data with minor issues |
Encoded in filename: `*_q10_*`, `*_q9_*`, `*_q8_*`.
---
## Caveats
Synthetic data may contain factual errors. `*_q8_*` files carry higher hallucination risk than `*_q10_*`.
---
## License
Apache 2.0
提供机构：
JohnTdi
