How agents and models behave inside simulated companies. Every result is graded from state, with its method and limits.
The ranking changed completely between easier and harder tasks.
Every model did worse on harder tasks, but not by the same amount.
Share of tries that passed, out of 50 per model on each set of tasks.