design blocks 105, completed 105, missing 0

=== reader claude-opus-5-5: 63 blocks, 21 problems, infra-touched blocks 0, P1 failed (not correct or <1x) 9 | metric score
arm means (mean over problems of per-problem rep means): bare 2.470, show_all 1.697, bare2 3.325, attempt_then_consult 3.030, consult_on_failure 3.269, audit_then_consult 3.125, prior_then_consult 3.304
  PRIMARY attempt_then_consult - bare2             -0.295 [-0.706, +0.130]  problems up/down 4/17
  audit_then_consult - bare2                       -0.199 [-0.541, +0.156]  problems up/down 7/14
  consult_on_failure - bare2                       -0.056 [-0.120, -0.006]  problems up/down 0/4
  attempt_then_consult - prior_then_consult        -0.275 [-0.579, -0.019]  problems up/down 6/15
  audit_then_consult - attempt_then_consult        +0.096 [-0.105, +0.285]  problems up/down 14/6
  prior_then_consult - bare2                       -0.020 [-0.401, +0.296]  problems up/down 8/13
  show_all - bare                                  -0.773 [-1.529, -0.191]  problems up/down 6/15
  bare2 - bare                                     +0.855 [+0.500, +1.261]  problems up/down 18/1
  attempt_then_consult - bare                      +0.560 [+0.276, +0.959]  problems up/down 16/4
  audit_then_consult - bare                        +0.656 [+0.380, +0.984]  problems up/down 18/1
  consult_on_failure - bare                        +0.799 [+0.467, +1.184]  problems up/down 18/1
  prior_then_consult - bare                        +0.835 [+0.500, +1.225]  problems up/down 19/2
  attempt_then_consult - show_all                  +1.333 [+0.653, +2.218]  problems up/down 20/1
  consult_on_failure - prior_then_consult          -0.035 [-0.347, +0.341]  problems up/down 10/11
  audit_then_consult - prior_then_consult          -0.179 [-0.443, +0.033]  problems up/down 10/11
  diag P2_consult - P2_refine (raw second pass)    -0.190 [-0.690, +0.275]  n_problems 21
  diag P2_audit - P2_refine (raw second pass)      -0.171 [-0.547, +0.189]  n_problems 21
  diag P2_prior - P2_refine (raw second pass)      +0.035 [-0.392, +0.397]  n_problems 21
  diag S - P1 (raw pass evals)                     -0.761 [-1.527, -0.172]  n_problems 21
  selection (which pass each two-pass arm kept): bare2: {'P1': 21, 'P2_refine': 42}; attempt_then_consult: {'P1': 19, 'P2_consult': 44}; audit_then_consult: {'P2_audit': 47, 'P1': 16}; prior_then_consult: {'P1': 14, 'P2_prior': 49}

=== reader claude-opus-4-8: 42 blocks, 21 problems, infra-touched blocks 0, P1 failed (not correct or <1x) 31 | metric score
arm means (mean over problems of per-problem rep means): bare 0.625, show_all 1.112, bare2 1.155, attempt_then_consult 1.304, consult_on_failure 1.311, audit_then_consult 1.322, prior_then_consult 1.249
  PRIMARY attempt_then_consult - bare2             +0.149 [-0.075, +0.504]  problems up/down 12/9
  audit_then_consult - bare2                       +0.167 [-0.045, +0.520]  problems up/down 13/8
  consult_on_failure - bare2                       +0.156 [-0.064, +0.516]  problems up/down 13/7
  attempt_then_consult - prior_then_consult        +0.055 [-0.061, +0.188]  problems up/down 12/9
  audit_then_consult - attempt_then_consult        +0.018 [-0.043, +0.098]  problems up/down 12/8
  prior_then_consult - bare2                       +0.094 [-0.038, +0.330]  problems up/down 9/12
  show_all - bare                                  +0.487 [+0.249, +0.716]  problems up/down 18/3
  bare2 - bare                                     +0.529 [+0.319, +0.793]  problems up/down 18/2
  attempt_then_consult - bare                      +0.679 [+0.398, +1.029]  problems up/down 20/1
  audit_then_consult - bare                        +0.697 [+0.389, +1.080]  problems up/down 20/1
  consult_on_failure - bare                        +0.685 [+0.400, +1.044]  problems up/down 20/1
  prior_then_consult - bare                        +0.623 [+0.368, +0.935]  problems up/down 18/1
  attempt_then_consult - show_all                  +0.192 [+0.021, +0.408]  problems up/down 11/9
  consult_on_failure - prior_then_consult          +0.062 [-0.050, +0.192]  problems up/down 13/8
  audit_then_consult - prior_then_consult          +0.073 [-0.024, +0.198]  problems up/down 14/7
  diag P2_consult - P2_refine (raw second pass)    +0.213 [-0.037, +0.583]  n_problems 21
  diag P2_audit - P2_refine (raw second pass)      +0.116 [-0.255, +0.523]  n_problems 21
  diag P2_prior - P2_refine (raw second pass)      +0.110 [-0.089, +0.380]  n_problems 21
  diag S - P1 (raw pass evals)                     +0.437 [+0.208, +0.666]  n_problems 21
  selection (which pass each two-pass arm kept): bare2: {'P1': 12, 'P2_refine': 30}; attempt_then_consult: {'P1': 8, 'P2_consult': 34}; audit_then_consult: {'P1': 10, 'P2_audit': 32}; prior_then_consult: {'P1': 16, 'P2_prior': 26}

(secondary diagnostic: metric score, wave pooled; preregistered decision rules apply only to --wave 1 --metric score and are not evaluated here)