PuzzleJAX: A Benchmark for Reasoning and Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Earle, Sam, Todd, Graham, Li, Yuchen, Khalifa, Ahmed, Nasir, Muhammad Umair, Jiang, Zehua, Banburski-Fahey, Andrzej, Togelius, Julian
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912551224213504
author Earle, Sam
Todd, Graham
Li, Yuchen
Khalifa, Ahmed
Nasir, Muhammad Umair
Jiang, Zehua
Banburski-Fahey, Andrzej
Togelius, Julian
author_facet Earle, Sam
Todd, Graham
Li, Yuchen
Khalifa, Ahmed
Nasir, Muhammad Umair
Jiang, Zehua
Banburski-Fahey, Andrzej
Togelius, Julian
contents We introduce PuzzleJAX, a GPU-accelerated puzzle game engine and description language designed to support rapid benchmarking of tree search, reinforcement learning, and LLM reasoning abilities. Unlike existing GPU-accelerated learning environments that provide hard-coded implementations of fixed sets of games, PuzzleJAX allows dynamic compilation of any game expressible in its domain-specific language (DSL). This DSL follows PuzzleScript, which is a popular and accessible online game engine for designing puzzle games. In this paper, we validate in PuzzleJAX several hundred of the thousands of games designed in PuzzleScript by both professional designers and casual creators since its release in 2013, thereby demonstrating PuzzleJAX's coverage of an expansive, expressive, and human-relevant space of tasks. By analyzing the performance of search, learning, and language models on these games, we show that PuzzleJAX can naturally express tasks that are both simple and intuitive to understand, yet often deeply challenging to master, requiring a combination of control, planning, and high-level insight.
format Preprint
id arxiv_https___arxiv_org_abs_2508_16821
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PuzzleJAX: A Benchmark for Reasoning and Learning
Earle, Sam
Todd, Graham
Li, Yuchen
Khalifa, Ahmed
Nasir, Muhammad Umair
Jiang, Zehua
Banburski-Fahey, Andrzej
Togelius, Julian
Artificial Intelligence
Machine Learning
We introduce PuzzleJAX, a GPU-accelerated puzzle game engine and description language designed to support rapid benchmarking of tree search, reinforcement learning, and LLM reasoning abilities. Unlike existing GPU-accelerated learning environments that provide hard-coded implementations of fixed sets of games, PuzzleJAX allows dynamic compilation of any game expressible in its domain-specific language (DSL). This DSL follows PuzzleScript, which is a popular and accessible online game engine for designing puzzle games. In this paper, we validate in PuzzleJAX several hundred of the thousands of games designed in PuzzleScript by both professional designers and casual creators since its release in 2013, thereby demonstrating PuzzleJAX's coverage of an expansive, expressive, and human-relevant space of tasks. By analyzing the performance of search, learning, and language models on these games, we show that PuzzleJAX can naturally express tasks that are both simple and intuitive to understand, yet often deeply challenging to master, requiring a combination of control, planning, and high-level insight.
title PuzzleJAX: A Benchmark for Reasoning and Learning
topic Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2508.16821