README.md

metadata

license: mit
task_categories:
  - text-generation
language:
  - en
tags:
  - undertale
  - gaming
  - character-roleplay
  - alpaca
  - instruction-tuning
  - fine-tuning
size_categories:
  - 10K<n<100K

Undertale Alpaca Dataset

Dataset Description

An Alpaca-format instruction dataset covering 11 characters from the indie RPG Undertale by Toby Fox.

Characters Included

Sans, Papyrus, Toriel, Asgore, Undyne, Alphys, Flowey, Mettaton, Napstablook, Chara, and Muffet.

Dataset Structure

Each entry follows the Stanford Alpaca format:

{
  "instruction": "Describe Sans's personality in detail.",
  "input": "",
  "output": "Sans from Undertale has the following personality traits: ..."
}

Per-Character Coverage (10 entries each)

1. Character Overview — Species, role, stats, personality summary
2. Personality — Detailed personality trait description
3. Appearance — Visual description of the character
4. Abilities — Powers and combat abilities
5. Backstory — Full character backstory and lore
6. Key Quotes — Iconic in-game lines with context
7. Relationships — Character's connections to others
8. Fun Facts — Trivia, dev notes, community knowledge
9. Battle Stats — HP, ATK, DEF and stat analysis
10. Roleplay Prompt — In-character introduction for RP fine-tuning

Stats

• Total entries: 110
• Characters: 11
• Format: Alpaca (JSONL)
• Language: English

Intended Use

Fine-tuning instruction-following models (LLaMA, Mistral, etc.) for Undertale character knowledge and roleplay using tools like Axolotl, LLaMA-Factory, or unsloth.