Reproduce “Recursive Game Creator: An Agentic Product-Level Experience-Oriented Game Harness” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jiajun Chen, Haoyu Wu, Mingda Jia, Xihui Liu
Recursive Game Creator uses repeated play to guide game revisions. Its Designer and Builder create a candidate; a coding-native Player runs policies through exposed game state and actions, and a Reviewer combines trajectories, evidence, and preferences into feedback. On 45 selected Godot tasks, GPT-6 Astra (high) reaches 77.89/100 after three rounds, versus 71.26 for its baseline; this does not establish human enjoyment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.