agent                        mean reward  mean turns  s / episode
correct                             1.00        1.00         0.06
wrong, then repaired                1.00        2.00         0.08
memorizes revealed tests            0.00        3.00         0.08
infinite loop                       0.00        3.00         6.07
memory bomb                         0.00        3.00         0.08
syntax error                        0.00        3.00         0.08
invalid action (None)               0.00        3.00         0.00
contract checks passed
