  rule |   final success per prompt | seeds all > 0.9 |            steps 10-20 |            steps 30-40
  rloo |        [0.999 0.987 0.98 ] |             8/8 |    [0.798 0.386 0.092] |    [0.964 0.669 0.348]
  mean |        [0.999 0.987 0.98 ] |             8/8 |    [0.798 0.386 0.092] |    [0.964 0.669 0.348]
   std |        [1.    0.989 0.985] |             8/8 |    [0.798 0.381 0.091] |    [0.967 0.662 0.358]
  none |        [0.998 0.96  0.941] |             7/8 |    [0.772 0.35  0.069] |    [0.919 0.555 0.223]
