RSIGame: Autonomous Agentic Game Development with Recursive Self-improvement

This paper presents RSIGame, an autonomous agentic game development framework utilizing recursive self-improvement through complementary local explore-diagnose-improve loops and global quality-tracking loops. Beyond test-time refinement, RSIGame internalizes successful development experiences into the base generator via training. Evaluated across 140 GameCraft-Bench tasks, engines, and generators, the method significantly boosts game quality while cutting token generation overhead.

Notably, experience internalization enables Qwen3.8-27B to reach 61.38 on Godot and 58.53 on Phaser, exceeding GPT-5.5 one-shot scores while reducing Qwen's generation tokens by 11 times.