language:

• en High Priest Catholicism Training Dataset tags:
• catholicism
• theology
• llm-training
• instruction-tuning
• religious-texts
• pastoral-care
• doctrinal-study

size_categories: - 100K<n<1M license: other

High Priest Catholicism Training Dataset

Dataset Summary

A 102,202-example instruction-response dataset for training LLMs to reason and think from the perspective of a Catholic High Priest. All content is fact-based, extracted from 34 authoritative Catholic texts with no placeholders and no duplicates.

Dataset Details

Basic Information

• Total Examples: 102,202
• Format: JSONL (one JSON object per line)
• File: high_priest_dataset.jsonl (58.3 MB)
• Languages: English
• Creation Date: May 2026

Key Features

✅ No placeholders - All content is fact-based from source texts
✅ No duplicates - Deduplicated via MD5 hashing
✅ Diverse categories - 9 theological/moral reasoning categories
✅ Authoritative sources - 34 official Catholic texts including Church Fathers, Vatican documents, and the Catechism

Sources

Extracted from 34 PDF books in R:\Data-Sets\high_priest_Catholicism\books\:

Dataset Structure

Each line is a JSON object with 4 fields:

{
 "instruction": "What does the Catholic Church teach about the Eucharist?",
 "response": "According to Catechism of the Catholic Church - USCCB.pdf: The Eucharist is the source and summit of the Christian life...",
 "category": "theological_qa",
 "source": "Catechism of the Catholic Church - USCCB.pdf"
}

Fields

Category Distribution

Usage

Load with Hugging Face datasets library:

from datasets import load_dataset

# Load from local JSONL file
dataset = load_dataset("json", data_files="high_priest_dataset.jsonl")

# Or load from Hugging Face Hub (after uploading)
# dataset = load_dataset("your-username/high-priest-catholicism")

# Split into train/validation
dataset = dataset["train"].train_test_split(test_size=0.1)
train = dataset["train"]
val = dataset["test"]

# Example usage
for example in train.select(range(3)):
 print(f"Instruction: {example['instruction']}")
 print(f"Response: {example['response'][:100]}...")
 print("---")

Intended Use

• Instruction-tuning: Train LLMs to adopt a Catholic High Priest mindset
• Theological Q&A: Generate doctrinally sound responses to Catholic questions
• Pastoral Training: Simulate priestly counsel and moral reasoning
• Research: Study Catholic theological discourse patterns

Limitations

• Responses are truncated for brevity (350-500 characters)
• All content is in English
• Limited to the 34 source texts provided (may not cover all niche topics)
• Some source PDFs had extraction limitations (e.g., scanned texts)

Ethical Considerations

• This dataset represents Catholic teaching; users should verify doctrine for critical applications
• Source texts are a mix of public domain and copyrighted materials - respect original copyrights
• The dataset trains a specific theological perspective; not suitable for neutral religious queries

How to Upload to Hugging Face Hub

# Install Hugging Face CLI
pip install huggingface_hub

# Login
huggingface-cli login

# Create dataset repository
huggingface-cli repo create high-priest-catholicism --type dataset

# Upload
huggingface-cli upload high-priest-catholicism high_priest_dataset.jsonl
huggingface-cli upload high-priest-catholicism README.md

Acknowledgments

Dataset generated from 34 Catholic texts including:

• Official Catechism of the Catholic Church
• Vatican II Documents
• Church Fathers (Augustine, Irenaeus, Athanasius, Anselm)
• Spiritual writers (St. John of the Cross, St. Teresa of Avila, Thomas à Kempis)
• Papal documents (Laudato Si', Evangelii Gaudium)
• Canon Law and liturgical texts

---

Dataset Location: R:\Data-Sets\high_priest_Catholicism\dataset\high_priest_dataset.jsonl