This research paper explores a novel method for personalized procedural content generation in gaming. Utilizing large language models, the authors propose game levels based on individual player’s gameplay data. The method aims to minimize the ‘cold start problem’ usually encountered in machine learning based methods which require large amounts of data and complex model training. The technique demonstrated better player retention compared to traditional procedural generation techniques and is feasible for production settings.
Publication date: 16 Feb 2024
Project Page: Not provided
Paper: https://arxiv.org/pdf/2402.10133
