Title:	Game Level GenAI
DNr:	LiU-gpu-2025-12
Project Type:	LiU Compute
Principal Investigator:	Aseel Berglund <aseel.berglund@liu.se>
Affiliation:	Linköpings universitet
Duration:	2025-10-13 – 2026-01-01
Classification:	10201
Keywords:

Abstract

This research project, Game Level GenAI, investigates the potential of LLMs for game level generation, addressing fundamental challenges in dataset requirements, cross-genre applicability, model efficiency, and real-time performance. The research builds upon recent advancements in LLM-based PCG while addressing critical limitations in current approaches related to project-specific optimization, consistent playability, and computational efficiency. The Game Level GenAI project aims to advance the state of the art in AI-assisted game level design by: 1. Developing novel techniques for fine-tuning and prompting LLMs to generate high-quality, playable game levels 2. Creating a framework for evaluating LLM-generated levels across multiple dimensions: playability, novelty, diversity, and difficulty