← All the evidence

sentinel-interruption-heldout-results.txt

SHA-256 e6463beab9fb97dae9a20c4635915e2f0f61d4d6f2c3435ab45ad2374e8308ca · download the original

completed heldout ordinary
completed heldout drift
completed heldout delayed
completed heldout rare
completed heldout miscalibrated
completed heldout biased_audit
completed heldout personal
completed heldout personal_drift

biased_audit
  silent                   recall   0.00 prec   None false/1000   0.00 util/1000    0.00 late    0.00 rare   0.00
  fixed                    recall  14.17 prec  65.35 false/1000  27.24 util/1000   60.05 late   59.52 rare  12.92
  naive_silence_negative   recall   2.16 prec  78.36 false/1000   2.16 util/1000   20.01 late    0.00 rare   1.82
  bandit_explore           recall  14.30 prec  64.85 false/1000  28.10 util/1000   54.81 late   50.25 rare  14.38
  audit_only               recall  14.83 prec  69.49 false/1000  23.61 util/1000   78.01 late   75.53 rare  12.01
  prior_paced              recall  16.36 prec  80.17 false/1000  14.67 util/1000  130.92 late  126.32 rare  11.10
  feedback_paced           recall  14.32 prec  75.71 false/1000  16.65 util/1000  131.67 late  146.51 rare  13.47
  audit_paced              recall  13.28 prec  74.11 false/1000  16.82 util/1000  117.37 late  129.64 rare  11.28
  audit_stratified         recall  14.66 prec  69.73 false/1000  23.07 util/1000   84.69 late   89.63 rare  17.93
  audit_plus_probe         recall  14.68 prec  68.73 false/1000  24.22 util/1000   75.71 late   77.62 rare  12.19
  oracle                   recall  22.06 prec  100.0 false/1000   0.00 util/1000  166.46 late  166.95 rare  24.39

delayed
  silent                   recall   0.00 prec   None false/1000   0.00 util/1000    0.00 late    0.00 rare   0.00
  fixed                    recall  14.27 prec  66.18 false/1000  26.59 util/1000   67.58 late   67.91 rare  14.65
  naive_silence_negative   recall   1.57 prec  81.48 false/1000   1.30 util/1000   13.81 late    0.00 rare   1.36
  bandit_explore           recall  14.24 prec  64.97 false/1000  27.98 util/1000   53.73 late   46.34 rare  12.65
  audit_only               recall  15.04 prec  72.07 false/1000  21.25 util/1000   94.82 late   95.94 rare  15.65
  prior_paced              recall  16.38 prec  80.79 false/1000  14.19 util/1000  137.75 late  140.84 rare  13.47
  feedback_paced           recall  15.28 prec  77.32 false/1000  16.34 util/1000  138.07 late  146.16 rare  13.19
  audit_paced              recall  13.16 prec  75.73 false/1000  15.38 util/1000  112.64 late  125.26 rare  16.65
  audit_stratified         recall  14.88 prec  71.21 false/1000  21.93 util/1000   89.59 late   88.32 rare  17.38
  audit_plus_probe         recall  14.90 prec  70.75 false/1000  22.45 util/1000   88.41 late   89.55 rare  15.10
  oracle                   recall  21.95 prec  100.0 false/1000   0.00 util/1000  162.22 late  154.84 rare  20.93

drift
  silent                   recall   0.00 prec   None false/1000   0.00 util/1000    0.00 late    0.00 rare   0.00
  fixed                    recall  14.57 prec  59.86 false/1000  31.61 util/1000   48.83 late   30.78 rare  13.42
  naive_silence_negative   recall   2.46 prec  81.55 false/1000   1.80 util/1000   20.26 late    0.00 rare   1.18
  bandit_explore           recall  15.37 prec   62.3 false/1000  30.09 util/1000   43.70 late   42.89 rare  12.15
  audit_only               recall  15.31 prec  69.12 false/1000  22.12 util/1000   75.32 late   86.48 rare  11.06
  prior_paced              recall  17.73 prec  77.39 false/1000  16.76 util/1000  111.54 late   90.22 rare  11.15
  feedback_paced           recall  15.09 prec  74.07 false/1000  17.08 util/1000  109.10 late  111.67 rare  12.96
  audit_paced              recall  13.23 prec  71.87 false/1000  16.74 util/1000   86.82 late   95.44 rare  10.52
  audit_stratified         recall  15.98 prec  69.27 false/1000  22.93 util/1000   81.66 late   93.43 rare  17.86
  audit_plus_probe         recall  15.22 prec  67.17 false/1000  24.06 util/1000   70.09 late   83.70 rare  12.33
  oracle                   recall  24.73 prec  100.0 false/1000   0.00 util/1000  160.35 late  157.40 rare  25.84

miscalibrated
  silent                   recall   0.00 prec   None false/1000   0.00 util/1000    0.00 late    0.00 rare   0.00
  fixed                    recall  12.48 prec  56.69 false/1000  34.21 util/1000   18.69 late  -19.13 rare  12.28
  naive_silence_negative   recall   2.42 prec  80.24 false/1000   2.14 util/1000   22.67 late    2.89 rare   5.35
  bandit_explore           recall  12.79 prec  58.76 false/1000  32.20 util/1000   24.47 late    7.15 rare  15.96
  audit_only               recall  12.99 prec  63.84 false/1000  26.41 util/1000   49.50 late   13.61 rare  17.46
  prior_paced              recall  14.90 prec   71.6 false/1000  21.20 util/1000   81.93 late   36.56 rare  12.11
  feedback_paced           recall  11.91 prec  67.94 false/1000  20.17 util/1000   74.80 late   37.42 rare  15.79
  audit_paced              recall  10.91 prec  65.52 false/1000  20.60 util/1000   59.70 late   23.87 rare  18.95
  audit_stratified         recall  13.17 prec  63.83 false/1000  26.76 util/1000   51.86 late   20.34 rare  23.60
  audit_plus_probe         recall  13.00 prec  62.51 false/1000  27.97 util/1000   43.75 late   10.46 rare  15.96
  oracle                   recall  22.30 prec  100.0 false/1000   0.00 util/1000  162.48 late  160.66 rare  21.75

ordinary
  silent                   recall   0.00 prec   None false/1000   0.00 util/1000    0.00 late    0.00 rare   0.00
  fixed                    recall  14.13 prec  65.38 false/1000  27.30 util/1000   62.17 late   65.13 rare  12.11
  naive_silence_negative   recall   2.13 prec  82.66 false/1000   1.63 util/1000   22.04 late    1.90 rare   5.08
  bandit_explore           recall  14.13 prec  64.83 false/1000  27.99 util/1000   50.34 late   56.71 rare  10.86
  audit_only               recall  14.37 prec  70.85 false/1000  21.59 util/1000   83.25 late   94.14 rare  14.16
  prior_paced              recall  16.35 prec  81.24 false/1000  13.78 util/1000  132.29 late  141.51 rare  10.60
  feedback_paced           recall  14.12 prec  76.19 false/1000  16.11 util/1000  126.38 late  144.21 rare   8.99
  audit_paced              recall  12.91 prec  74.89 false/1000  15.81 util/1000  109.07 late  137.96 rare  14.78
  audit_stratified         recall  14.73 prec  70.58 false/1000  22.42 util/1000   86.96 late   95.39 rare  17.01
  audit_plus_probe         recall  14.32 prec  69.05 false/1000  23.44 util/1000   77.84 late   88.82 rare  14.34
  oracle                   recall  21.91 prec  100.0 false/1000   0.00 util/1000  161.41 late  161.16 rare  20.39

personal
  silent                   recall   0.00 prec   None false/1000   0.00 util/1000    0.00 late    0.00 rare   0.00
  fixed                    recall  14.30 prec  47.71 false/1000  41.17 util/1000   30.41 late   27.66 rare  13.69
  naive_silence_negative   recall   5.99 prec  66.39 false/1000   7.97 util/1000   85.33 late   81.15 rare   6.06
  bandit_explore           recall  13.07 prec  47.62 false/1000  37.79 util/1000   31.97 late   39.14 rare  10.85
  audit_only               recall  13.74 prec  59.07 false/1000  25.01 util/1000   90.09 late   93.43 rare   9.78
  prior_paced              recall  14.87 prec  53.16 false/1000  34.43 util/1000   89.11 late  103.41 rare  12.12
  feedback_paced           recall  14.56 prec  64.52 false/1000  21.04 util/1000  173.30 late  210.10 rare  12.51
  audit_paced              recall  12.40 prec  65.04 false/1000  17.51 util/1000  145.58 late  184.44 rare   9.58
  audit_stratified         recall  14.06 prec  57.58 false/1000  27.23 util/1000   93.86 late   97.04 rare  17.40
  audit_plus_probe         recall  13.77 prec  57.28 false/1000  26.99 util/1000   84.19 late   87.31 rare  11.63
  oracle                   recall  28.96 prec  100.0 false/1000   0.00 util/1000  248.46 late  254.97 rare  38.22

personal_drift
  silent                   recall   0.00 prec   None false/1000   0.00 util/1000    0.00 late    0.00 rare   0.00
  fixed                    recall  14.17 prec  52.44 false/1000  37.43 util/1000   56.88 late   69.88 rare  14.72
  naive_silence_negative   recall   5.71 prec   62.7 false/1000   9.88 util/1000   82.84 late   50.76 rare   4.45
  bandit_explore           recall  12.71 prec   48.4 false/1000  39.47 util/1000   34.76 late   40.82 rare  10.87
  audit_only               recall  13.97 prec  57.74 false/1000  29.78 util/1000  100.49 late   97.78 rare  10.87
  prior_paced              recall  14.69 prec  57.49 false/1000  31.63 util/1000  127.01 late  156.26 rare  15.12
  feedback_paced           recall  13.23 prec   62.1 false/1000  23.50 util/1000  147.43 late  148.42 rare  12.94
  audit_paced              recall  12.76 prec  64.09 false/1000  20.82 util/1000  146.45 late  169.37 rare  11.46
  audit_stratified         recall  13.81 prec  56.69 false/1000  30.73 util/1000   91.90 late   98.00 rare  16.50
  audit_plus_probe         recall  13.75 prec  55.89 false/1000  31.61 util/1000   90.58 late   85.78 rare  10.97
  oracle                   recall  26.86 prec  100.0 false/1000   0.00 util/1000  284.35 late  249.34 rare  32.11

rare
  silent                   recall   0.00 prec   None false/1000   0.00 util/1000    0.00 late    0.00 rare   0.00
  fixed                    recall  14.41 prec  66.07 false/1000  26.81 util/1000   61.12 late   63.22 rare  15.21
  naive_silence_negative   recall   2.43 prec  80.26 false/1000   2.16 util/1000   23.43 late    0.00 rare   2.30
  bandit_explore           recall  14.31 prec  64.89 false/1000  28.06 util/1000   51.94 late   53.16 rare  10.37
  audit_only               recall  15.29 prec  71.31 false/1000  22.30 util/1000   87.95 late   94.78 rare  10.83
  prior_paced              recall  16.68 prec   80.9 false/1000  14.27 util/1000  129.92 late  131.64 rare  11.98
  feedback_paced           recall  14.49 prec  76.13 false/1000  16.47 util/1000  130.10 late  132.77 rare   5.07
  audit_paced              recall  13.27 prec  75.17 false/1000  15.89 util/1000  109.75 late  118.84 rare   7.37
  audit_stratified         recall  14.78 prec  70.16 false/1000  22.77 util/1000   80.04 late   79.03 rare  13.59
  audit_plus_probe         recall  15.13 prec  70.09 false/1000  23.40 util/1000   81.68 late   85.23 rare  10.83
  oracle                   recall  22.07 prec  100.0 false/1000   0.00 util/1000  162.21 late  165.32 rare  20.51