license: cc-by-sa-4.0
language: en
multilinguality: monolingual
size_categories: n<1K
source_datasets:
- original
annotations_creators:
- machine-generated
language_creators:
- machine-generated
tags:
- rpg
- game-ai
- tactical-decision-making
- fantasy
- mud
- llm
- conversations
- turn-based-combat
- multi-turn
- simulation
- synthetic
- dialogue
- conversational-ai
- large-language-models
- instruction
- llm-training
- conversational-ai
- training-data
- retrieval-augmented-generation
- agentic
task_categories:
- text-generation
task_ids:
- dialogue-generation
- text-simplification
- language-modeling
name: RPG_DM_Simulation_Combat_LLM_Training
pretty_name: Magician MUD Conversations
description: 20 turn-by-turn gameplay conversations from a text-based dungeon crawler RPG (MUD style). Each conversation captures strategic decision-making in fantasy combat, including player status, enemy encounters, resource management, and combat outcomes. Ideal for fine-tuning language models for RPG dialogue generation, tactical decision-making, and game state understanding.
Get the full 30K record dataset on Gumroad at https://datadeveloper1.gumroad.com/l/lmfhbg
The full CJ Jones' synthetic dataset catalog is available at: https://datadeveloper1.gumroad.com
Want more? 🚀 Get the AI Startup Bundle from Gumroad.
dataset_size: 500 splits: train: 400 validation: 50 test: 50 dataset_structure: description: Each instance represents a single conversation turn. fields: conversation_id: string game_id: string turn_number: int speaker: string message: string game_state: dict selected_choice: string choice_number: int choice_reason: string attacked_entities: list combat_outcome: dict game_outcome: string considerations: social_impact: Fantasy violence only; no real-world sensitive content. bias: > Contains combat-focused scenarios in a Western fantasy RPG setting. AI choices may not reflect human player behavior. limitations: > Small dataset (500 conversations). Synthetic data, domain-specific. Not suitable for large-scale pre-training. recommended_use_cases:
- Fine-tuning small to medium language models (≤7B parameters)
- Training supervised game-playing agents
- RPG dialogue systems
- Tactical AI research
- Game design education not_recommended_use_cases:
- Large-scale pre-training
- Real-world decision-making systems
- Medical, financial, or safety-critical applications citation: bibtex: | @dataset{magician-mud-conversations-2025, title = {Magician MUD Conversations: A Dataset of 500 Tactical RPG Dialogues}, author = {Magician MUD Simulator Team}, year = {2025}, publisher = {Hugging Face}, version = {1.0.0}, url = {https://huggingface.co/datasets/RPG_DM_Simulation_Combat_LLM_Training} }