awgym(runtime, public)
An ARC training gym — a game a world model can watch, and six roles that play through it.
Name
awgym — An ARC training gym — a game a world model can watch, and six roles that play through it.
The problem it solves
A world model is only as good as the play it watches. Solver runs stream past, checkpoints get trained on whatever happened to be recorded, and nobody can answer "did the model actually get better at predicting the next grid?" against transitions it has never seen.
Instead of trusting / you check
- instead of trusting
- that a world model learned anything from the games it saw
- you check
- transitions captured from real play, fed back, and the retrodiction score falling on grids it never saw
Adopt it alone
Play one ARC game end-to-end with the world model watching every transition, then ask it to predict the next grid and watch the surprise score fall.
Install
pip install awgym
Where it sits in the path
- Not on the path yet — reference only.
Links
Generated 2026-09-14 00:48Z from the ecosystem registry. A field here that is wrong is wrong in the registry — fix it there and this page follows.