Prompt-Guided Level Generation

Abstract

Automated generation of complex and diverse environments can be achieved through the use of Procedural Content Generation (PCG) algorithms. However, generating content that is both meaningful and reflective of specific intentions and constraints remains a challenge. Recent advances in Large Language Models (LLMs) have demonstrated their effectiveness in various domains. These models can be fine-tuned and information can be reused to accelerate training for new tasks. Our study presents MarioGPT, a fine-tuned GPT2 model that has been trained to generate tile-based game levels for Super Mario Bros. The results demonstrate that MarioGPT can generate diverse levels and can be text-prompted for controllable level generation, addressing a critical challenge in current PCG techniques.