Practical PCG Through Large Language Models

Muhammad U. Nasir; Julian Togelius

doi:10.1109/CoG57401.2023.10333197

Practical PCG Through Large Language Models

Muhammad U. Nasir, Julian Togelius

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Large Language Models (LLMs) have proven to be useful tools in various domains outside of the field of their inception, which was natural language processing. In this study, we provide practical directions on how to use LLMs to generate 2D-game rooms for an under-development game, named Metavoidal. Our technique can harness the power of GPT-3 by Human-in-the-loop fine-tuning which allows our method to create 37% Playable-Novel levels from as scarce data as only 60 hand-designed rooms under a scenario of the non-trivial game, with respect to (Procedural Content Generation) PCG, that has a good amount of local and global constraints.

Original language	English (US)
Title of host publication	Proceedings of the 2023 IEEE Conference on Games, CoG 2023
Publisher	IEEE Computer Society
ISBN (Electronic)	9798350322774
DOIs	https://doi.org/10.1109/CoG57401.2023.10333197
State	Published - 2023
Event	5th Annual IEEE Conference on Games, CoG 2023 - Boston, United States Duration: Aug 21 2023 → Aug 24 2023

Publication series

Name	IEEE Conference on Computatonal Intelligence and Games, CIG
ISSN (Print)	2325-4270
ISSN (Electronic)	2325-4289

Conference

Conference	5th Annual IEEE Conference on Games, CoG 2023
Country/Territory	United States
City	Boston
Period	8/21/23 → 8/24/23

Keywords

Large Language Models
Procedural Content Generation

ASJC Scopus subject areas

Artificial Intelligence
Computer Graphics and Computer-Aided Design
Computer Vision and Pattern Recognition
Human-Computer Interaction
Software

Access to Document

10.1109/CoG57401.2023.10333197

Cite this

@inproceedings{92aaf10f2849499b9ab873f20c56f324,

title = "Practical PCG Through Large Language Models",

abstract = "Large Language Models (LLMs) have proven to be useful tools in various domains outside of the field of their inception, which was natural language processing. In this study, we provide practical directions on how to use LLMs to generate 2D-game rooms for an under-development game, named Metavoidal. Our technique can harness the power of GPT-3 by Human-in-the-loop fine-tuning which allows our method to create 37% Playable-Novel levels from as scarce data as only 60 hand-designed rooms under a scenario of the non-trivial game, with respect to (Procedural Content Generation) PCG, that has a good amount of local and global constraints.",

keywords = "Large Language Models, Procedural Content Generation",

author = "Nasir, {Muhammad U.} and Julian Togelius",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; 5th Annual IEEE Conference on Games, CoG 2023 ; Conference date: 21-08-2023 Through 24-08-2023",

year = "2023",

doi = "10.1109/CoG57401.2023.10333197",

language = "English (US)",

series = "IEEE Conference on Computatonal Intelligence and Games, CIG",

publisher = "IEEE Computer Society",

booktitle = "Proceedings of the 2023 IEEE Conference on Games, CoG 2023",

}

TY - GEN

T1 - Practical PCG Through Large Language Models

AU - Nasir, Muhammad U.

AU - Togelius, Julian

PY - 2023

Y1 - 2023

N2 - Large Language Models (LLMs) have proven to be useful tools in various domains outside of the field of their inception, which was natural language processing. In this study, we provide practical directions on how to use LLMs to generate 2D-game rooms for an under-development game, named Metavoidal. Our technique can harness the power of GPT-3 by Human-in-the-loop fine-tuning which allows our method to create 37% Playable-Novel levels from as scarce data as only 60 hand-designed rooms under a scenario of the non-trivial game, with respect to (Procedural Content Generation) PCG, that has a good amount of local and global constraints.

AB - Large Language Models (LLMs) have proven to be useful tools in various domains outside of the field of their inception, which was natural language processing. In this study, we provide practical directions on how to use LLMs to generate 2D-game rooms for an under-development game, named Metavoidal. Our technique can harness the power of GPT-3 by Human-in-the-loop fine-tuning which allows our method to create 37% Playable-Novel levels from as scarce data as only 60 hand-designed rooms under a scenario of the non-trivial game, with respect to (Procedural Content Generation) PCG, that has a good amount of local and global constraints.

KW - Large Language Models

KW - Procedural Content Generation

UR - http://www.scopus.com/inward/record.url?scp=85180546375&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85180546375&partnerID=8YFLogxK

U2 - 10.1109/CoG57401.2023.10333197

DO - 10.1109/CoG57401.2023.10333197

M3 - Conference contribution

AN - SCOPUS:85180546375

T3 - IEEE Conference on Computatonal Intelligence and Games, CIG

BT - Proceedings of the 2023 IEEE Conference on Games, CoG 2023

PB - IEEE Computer Society

T2 - 5th Annual IEEE Conference on Games, CoG 2023

Y2 - 21 August 2023 through 24 August 2023

ER -

Practical PCG Through Large Language Models

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this