arXiv cs.AI· Sam Earle, Graham Todd, Yuchen Li, Ahmed Khalifa, Muhammad Umair Nasir, Zehua Jiang, Andrzej Banburski-Fahey, Julian Togelius·· 10 天前AI 评分27
PuzzleJAX：面向推理与学习的基准测试
PuzzleJAX: A Benchmark for Reasoning and Learning
AI 导读
PuzzleJAX 是一个 GPU 加速的益智游戏引擎与描述语言，用于快速评测树搜索、强化学习和 LLM 推理能力。它基于 PuzzleScript 的 DSL 支持动态编译任意可表达的游戏，并验证了 PuzzleScript 自 2013 年以来数百款游戏。论文发表于 IEEE Conference on Games 2026。
正文 · 原文
Abstract: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.
| Comments: | 25 pages, 11 figures, 2 tables, published as a full paper at IEEE Conference on Games 2026 |
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2508.16821 [cs.AI] |
| (or arXiv:2508.16821v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2508.16821 arXiv-issued DOI via DataCite |
Submission history
From: Sam Earle [view email]
[v1]
Fri, 22 Aug 2025 22:40:58 UTC (1,400 KB)
[v2]
Tue, 6 Oct 2026 01:51:16 UTC (2,263 KB)
来源：arXiv cs.AI · arxiv.org