●builderYou can use this to measure how well your coding agents actually handle complex engine APIs and serialized states.
●researcherThis provides a more rigorous, deterministic evaluation for agents operating in 3D environments compared to text-based benchmarks.