Model and harness configurations tested on realistic product management jobs, graded on structured checks and reviewed blind by me. Pick the job you have; I’ll show you what to use.
0 models0 configurations8 core tasks0 reviewed runs
A configuration earns an overall rank only once it has a graded, blind-reviewed run on every one of the 8 core tasks. Until then, provisional results appear in the ratings below.
The leading configuration in each category. Run counts are per category, so evidence is thinner here.
Only configurations that completed every core task get a rank. Scores out of 100.
No published results yet. The first release is being graded and reviewed.