---
pretty_name: ASFT
language:
- en
license: other
---

# ASFT

## Dataset Summary
ASFT is a collection of instruction–response pairs in JSONL format, intended for instruction tuning and general text-generation experiments. Each line is a single JSON object with two fields: `instruction` and `response`. Some files are derived from MedMCQA and include explanations.

## Files
- `magpie10k.jsonl`
- `numina_cot_10k.jsonl`
- `numina_cot_30k.jsonl`
- `numina_cot_100k.jsonl`
- `train_medmcqa_alpaca_10k.jsonl`
- `train_medmcqa_alpaca_30k.jsonl`
- `train_medmcqa_alpaca_100k.jsonl`

## Data Format
Each line in every file is a JSON object with the following schema:

```json
{
  "instruction": "...",
  "response": "..."
}
```

## Intended Use
- Instruction tuning / supervised fine-tuning
- General-purpose text generation benchmarks

## Data Provenance
- `train_medmcqa_alpaca_10k.jsonl`, `train_medmcqa_alpaca_30k.jsonl`, and `train_medmcqa_alpaca_100k.jsonl` are derived from the `openlifescienceai/medmcqa` training split (examples with non-empty explanations). `instruction` contains the question with options; `response` contains the explanation followed by the final answer sentence.
- The original data sources and collection process for the remaining files are not specified in this repository.

## Licensing
License is not specified. Please verify licensing and usage constraints before using this dataset in production or redistribution.

## Ethical Considerations
- The dataset has not been manually reviewed for sensitive or personal information.
- Use appropriate filtering and validation for downstream applications.

## Citation
If you use this dataset, please cite this Hugging Face repository.