Datasets:

pthinc
/

turkish_instruct_dataset_micro

license: mit
task_categories:
  - text-generation
  - question-answering
language:
  - tr
tags:
  - instruction-following
  - synthetic
  - turkish
size_categories:
  - 10K<n<100K
pretty_name: Turkish Instruct PTH

Turkish Instruct Dataset (Micro)

Dataset Description

This dataset contains 10169 synthetic samples generated using AI models. It is designed for fine-tuning Turkish LLMs on instruction-following, Q&A, Refusal, Chat, and Logic tasks.

Statistics

Total Samples: 10169
Categories: Instruction, Q&A, Refusal, Chat, Logic

Files

turkish_dataset_final_20k.csv: Raw CSV data
turkish_dataset_final_20k.json: JSON format
turkish_dataset_final_20k.parquet: Efficient Parquet format