A2Z GameSpec-Bench
What it is
A benchmark of 100 long-form Game Design Documents (GDDs) designed to evaluate how faithfully coding agents can generate complete games. It converts GDDs into dependency-aware contracts and uses source-code inspection combined with agent-generated test policies to measure requirement satisfaction and interdependencies.
The problem it solves
Existing game-development benchmarks typically use compact specifications and fail to evaluate the complex, interdependent requirements found in long-form Game Design Documents, making it difficult to assess if agents truly preserve intended design.
What you could use it for
- Evaluating the end-to-end specification-following ability of coding agents
- Benchmarking the ability of LLMs to satisfy interdependent game logic and visual requirements
- Providing targeted feedback to improve game development fidelity through self-revision
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.