Article Complete
AIM
Researchers Make AI-Generated Board Games Using CodeLLaMa
Researchers Make AI-Generated Board Games Using CodeLLaMa
Game Over, Humans

Researchers Make AI-Generated Board Games Using CodeLLaMa

GAVEL essentially uses a a readily available dataset of games that were made using the Ludii game system, generating variations of the games that don’t exist.

Published Updated

Key takeaways

  • Researchers use AI to create new board games through evolutionary computation and language models.
  • GAVEL generates game variations using the Ludii game system dataset, enhancing game diversity.
  • The innovative method combines game mechanics expressed as code with AI-driven mutations and recombinations.
  • One generated game, Havabu, merges elements from existing games Havannah and Tabu Y.
5 min

In a recent breakthrough, researchers have used the help of large language models (LLMs) to generate new and never-before-seen board games.

The researchers, from NYU, Maastricht University, Flinders University and UCLouvain, combined evolutionary computation with LLMs to create novel board games and their rules.

In the paper, titled ‘GAVEL: Generating Games Via Evolution and Language Models’,  the researchers made use of evolutionary computation to generate new board games based on the Ludii game generation language. Additionally, they made use of CodeLlama-13b to generate the rules of the game.

GAVEL essentially uses a readily available dataset of games that were made using the Ludii game system, generating variations of the games that don’t exist. This was done using an evolutionary computation method called MAP Elites, a population-based quality-diversity algorithm.

“We explore the generation of novel games in the comparatively expansive Ludii game description language, which encodes the rules of over 1000 board games in a variety of styles and modes of play.

We draw inspiration from recent advances in large language models and evolutionary computation in order to train a model that intelligently mutates and recombines games and mechanics expressed as code,” the researchers stated.

The games generated using the GAVEL method can be accessed through the Ludii portal. One game that the researchers generated is called Havabu, which combines elements of two Ludii games called Havannah and Tabu Y.

According to the researchers it’s “a variant of Havannah (a game in which players attempt to form loops or connected lines of pieces between sides of the board) that introduces a restriction on piece placement from another game in the Ludii dataset (Tabu Y).”

While novel, the use of generative AI is not uncommon in game development. There have been cases of indie developers using GenAI to develop their own games. However, these are used for specifics aspects of game creation, like creating art or assets, dialogue, or even music for the game. 

Membership

Stay ahead with AI insights.

Unlimited access to premium articles, exclusive analysis, and the stories shaping India's AI economy.

Subscribe now

AIM Americas

1190 Miraloma Way Suite P, Sunnyvale, CA 94085

AIM India

1st Floor, Sakti Statesman, Bellandur, Bengaluru – 560103

© 2026 AIM. All rights reserved.