initial:                    mass 0.389  H(correct) 1.313  rarest 0.00020  distinct@64 4.66  rarest@64 0.0127

training until pass@1 >= 0.9
  natural flow:             mass 0.900  H(correct) 1.313  rarest 0.00046  distinct@64 5.75  rarest@64 0.0292
  vanilla flow:             mass 0.900  H(correct) 0.275  rarest 0.00003  distinct@64 2.80  rarest@64 0.0021
  sampled REINFORCE (G=16): mass 0.903  H(correct) 0.914  rarest 0.00005  distinct@64 3.95  rarest@64 0.0030
      seed sd:  H(correct) 0.048  rarest 0.00001  distinct@64 0.19

training until pass@1 >= 0.99
  natural flow:             mass 0.990  H(correct) 1.313  rarest 0.00051  distinct@64 5.87  rarest@64 0.0321
  vanilla flow:             mass 0.990  H(correct) 0.038  rarest 0.00000  distinct@64 1.31  rarest@64 0.0002
  sampled REINFORCE (G=16): mass 0.990  H(correct) 0.564  rarest 0.00001  distinct@64 2.96  rarest@64 0.0004
      seed sd:  H(correct) 0.152  rarest 0.00000  distinct@64 0.15
