Datasetpublic

Kenshiii/prompt-templates

Prompt Templates Dataset Overview This dataset contains 300 unique prompt templates designed for use with large language models (LLMs). Each prompt is structured as a JSON object, making it easy to integrate into machine learning pipelines, especially those using the HuggingFace ecosystem. Features 300 unique prompts across various categories and difficulty levels. Structured JSONL format for easy ingestion into datasets. Diverse categories… See the full description on the dataset page: https://huggingface.co/datasets/Kenshiii/prompt-templates.

sourceHugging Facemitupdated 1y agoView on Hugging Face

0likes9downloads

Dataset Card

Prompt Templates Dataset

Overview

This dataset contains 300 unique prompt templates designed for use with large language models (LLMs). Each prompt is structured as a JSON object, making it easy to integrate into machine learning pipelines, especially those using the HuggingFace ecosystem.

Features

—300 unique prompts across various categories and difficulty levels.
—Structured JSONL format for easy ingestion into datasets.
—Diverse categories including creative writing, coding, Q&A, summarization, translation, reasoning, data wrangling, analysis, education, and more.
—English language only for consistency.
—Detailed instructions and expected output formats for each prompt.

Dataset Structure

Each entry in the dataset is a JSON object with the following fields:

Field	Description
`id`	Unique identifier for the prompt template (e.g., `pt-0001`).
`prompt`	The raw instruction to give an LLM.
`instruction`	A short human-readable summary of the task.
`input`	Example input or `null` if not applicable.
`expected_output_format`	Expected format of the output (e.g., `json`, `short_answer`, `bullet_list`).
`category`	Category of the prompt (e.g., `creative`, `coding`, `qa`).
`difficulty`	Difficulty level: `easy`, `medium`, or `hard`.
`strategy`	Prompting strategy used (e.g., `zero-shot`, `few-shot`, `CoT`).
`tags`	Topical tags related to the prompt.
`language`	ISO language code (e.g., `en` for English).
`notes`	Optional notes about the prompt (e.g., constraints, special instructions).

Usage

To use this dataset, simply load it into your preferred data processing tool or HuggingFace dataset loader:

python

from datasets import load_dataset

dataset = load_dataset('path/to/prompt_templates_en.jsonl')

Examples

Here are a few examples of the prompts included in the dataset:

1.Creative Writing

json

   {{
     "id": "pt-0001",
     "prompt": "Compose a medium-level short story centered around the theme of adventure in a futuristic city. Ensure the narrative includes at least a surprising plot twist. The final piece should be engaging and suitable for young adults.",
     "instruction": "Create a medium-level short story with a specific theme and elements.",
     "input": null,
     "expected_output_format": "short_paragraphs",
     "category": "creative",
     "difficulty": "medium",
     "strategy": "zero-shot",
     "tags": ["storytelling", "adventure"],
     "language": "en",
     "notes": "medium-level creative task"
   }}

1.Coding

json

   {{
     "id": "pt-0002",
     "prompt": "Develop a hard-level Python script that solves the problem of sorting a large dataset efficiently. Your solution should include detailed comments explaining each step and handle edge cases such as empty input. Provide a sample input and the expected output in your response.",
     "instruction": "Write and explain a hard-level Python script for a specific task.",
     "input": null,
     "expected_output_format": "code",
     "category": "coding",
     "difficulty": "hard",
     "strategy": "few-shot",
     "tags": ["python", "algorithm"],
     "language": "en",
     "notes": "hard-level coding task"
   }}

License

This dataset is provided under the MIT License. You are free to use, modify, and distribute it as needed.

Contributing

Contributions are welcome! If you have suggestions for new prompts, improvements to existing ones, or any other ideas, please open an issue or submit a pull request.