Uunan/turkish-planning-sft

Turkish Planning SFT

Turkish Planning SFT is a large-scale synthetic instruction-following dataset designed to improve the planning capabilities of Turkish Large Language Models (LLMs).

Rather than focusing on factual question answering, the dataset teaches models how to transform user goals, requirements, and constraints into structured, practical, and actionable plans.

The dataset is intended for Supervised Fine-Tuning (SFT) and follows a conversation-oriented format compatible with modern chat models.

Features

• —🇹🇷 Entirely in Turkish
• —🤖 Synthetic instruction-following dataset
• —📋 Planning-oriented conversations
• —🎯 Goal-driven tasks
• —🧩 Diverse real-world scenarios
• —📚 Multi-domain coverage
• —🛣️ Long-form structured responses
• —💬 Chat-based data format
• —⚙️ Compatible with modern SFT pipelines

Domains

The dataset covers a broad range of planning scenarios, including:

• —Business
• —Entrepreneurship
• —Education
• —Career Development
• —Software Engineering
• —Artificial Intelligence
• —Marketing
• —Finance
• —Productivity
• —Personal Development
• —Project Management
• —Healthcare
• —Human Resources
• —Creative Work
• —Research
• —Operations
• —Customer Service
• —Manufacturing
• —Agriculture
• —Non-profit Organizations

Dataset Format

Each sample is represented as a conversation.

[
  {
    "role": "user",
    "content": "...",
    "thinking": null,
    "images": null,
    "tool_calls": null
  },
  {
    "role": "assistant",
    "content": "...",
    "thinking": "...",
    "images": null,
    "tool_calls": null
  }
]

Fields

About the thinking Field

The thinking field is NOT chain-of-thought or hidden reasoning.

Instead, it contains a high-level planning abstraction describing aspects such as:

• —Understanding the user's objective
• —Identifying goals
• —Recognizing constraints
• —Considering potential risks
• —Selecting an appropriate planning strategy
• —Organizing the response
• —Choosing an appropriate communication style

It does not expose internal reasoning processes or intermediate inference steps.

Response Styles

Assistant responses include structured planning formats such as:

• —Roadmaps
• —Timelines
• —Checklists
• —Action Plans
• —Execution Plans
• —Milestone Plans
• —Weekly Plans
• —Monthly Plans
• —Strategic Plans
• —Learning Plans
• —Goal Decomposition
• —Improvement Plans
• —Implementation Strategies

Intended Use

This dataset is suitable for:

• —Supervised Fine-Tuning (SFT)
• —Turkish Instruction Tuning
• —Planning Assistants
• —Project Management Assistants
• —Productivity Assistants
• —Educational Assistants
• —Goal-Oriented AI Systems
• —Research on Planning Capabilities in Turkish LLMs

Generation

The dataset was synthetically generated using a hierarchical scenario generation pipeline.

The generation process combines:

• —Hierarchical topic selection
• —Diverse planning domains
• —Multiple personas
• —Variable constraints
• —Diverse planning templates
• —Structured conversational formatting

The objective is to maximize planning diversity while maintaining consistent instruction-following behavior.

Limitations

• —Entirely synthetic.
• —Designed specifically for planning and instruction-following tasks.
• —Does not aim to serve as a factual knowledge benchmark.
• —Responses should not replace professional advice in areas such as medicine, law, or finance.

Citation

If you use this dataset in your research, please cite:

@dataset{colak2026turkishplanningsft,
  title={Turkish Planning SFT},
  author={Uğurhan Çolak},
  year={2026},
  publisher={Hugging Face}
}

Author

Uğurhan Çolak

• —🤗 Hugging Face: https://huggingface.co/Uunan
• —💻 GitHub: https://github.com/Uunan

License

This dataset is distributed under the Apache-2.0 License.