Rebiha

Home › Datasets › Gaming & Game Master

Gaming & Game Master Fine-Tuning Dataset

NPC dialogue, quest design, narration, and RPG game mastering.

Fine-tune a model on this dataset

$19 for the dataset. Training is priced separately by model, starting at $47.

What is in this dataset

773 instruction-response pairs covering npc dialogue, quest design, narration, and rpg game mastering. Each pair is a realistic question with a full written answer, in the JSONL format used for supervised fine-tuning. No synthetic filler, no duplicated rows.

Example pairs

Real rows from the dataset, shown in full.

Write dialogue for a tavern keeper who secretly runs a thieves' guild from the back room. The players are new to town and ask for rumors.

Barkeep Tormund leans on the scarred oak counter, wiping a mug with a rag that's seen better days. His eyes flicker with a mix of warmth and calculation. 'Rumors, eh? Well, the road's been quiet, but they say old man Heston's barn burned down last night—strange, since he's been dead three winters. If you're looking for work, the notice board's by the door. But if you're looking for *real* work,' he lowers his voice, leaning closer, 'ask for the one who mends broken locks. He's in the back, through the cellar door, but only after the moon's high. And don't mention my name—I'm just the one who pours the ale.' His smile…

Design a side quest with a twist where the supposed villain is actually the victim of a misunderstanding.

Quest: 'The Crying Hollow' – The town of Thornwood hires the players to eliminate a 'monstrous beast' that's been stealing livestock from remote farms. The trail leads to a secluded cave in the Whispering Woods. Inside, they find not a monster, but a wounded, elderly griffin with a broken wing and a nest of eggs. The 'livestock' were actually a few sheep brought by a shepherd who stumbled upon the cave—the griffin only defended itself. The twist: the shepherd is the town's elder, spreading the rumor to cover his own greed (he wanted the griffin's feathers for a lucrative potion trade). The players must choose: expose the elder, help the griffin…

Narrate a combat encounter dynamically as a game master, focusing on a player's critical hit against a giant spider.

As the giant spider rears onto its bristling hind legs, its eight eyes glint with malice, venom dripping from its fangs. The paladin, Edric, charges forward, his longsword catching the dim light of the cavern's glowing fungi. He swings with a roar—and the dice clatter to a natural 20. 'You feel the blade bite deep into the chitinous hide, just below the spider's bulbous abdomen. Black ichor sprays across your shield as the creature shrieks—a high, piercing wail that echoes off the stone. The spider's legs buckle, and it skids sideways, leaving a trail of viscous fluid. For a moment, it's stunned, its mandibles twitching. Roll damage—and describe how you follow…

Who this is for

Game masters, NPC dialogue systems, and RPG companion tools.

Which models this works with

The dataset is plain JSONL, so it works with any instruction-tuned open model. On Rebiha you can train it directly on:

Training uses LoRA, or QLoRA on a 4-bit base for larger models. Your base model's weights are never modified — training produces a separate adapter.

How it works

  1. Pick this dataset and a base model.
  2. Rebiha runs the job on a managed GPU. You set up nothing.
  3. Download your model: a ready-to-run GGUF, plus developer assets (adapter, tokenizer, configs).

Fine-tuning or RAG?

Use RAG when the model needs facts that change — your prices, your documents, today's inventory. Use fine-tuning when you want consistent style and structure without prompting for it every time. This dataset teaches the shape of a good answer in this domain, not facts to look up.

Related datasets

Start fine-tuning