Get the benchmark
Download
Grab the SWE-game-bench harness to run the benchmark yourself, and the dataset of Unity game bugs used for evaluation.
SWE-game-bench harness
The full benchmark and evaluation harness, packaged as a .zip so anyone can download and run it.
Dataset (Zenodo)
The archived dataset of Unity game issues with their injected tests, hosted on Zenodo with a citable DOI.
Once you have results, submit them on the
Upload Results page.