=== MCB + Holm (authors' repro_stats.top_tier, B=10000, seed=0) ===

severe (any-severe indicator), clinical cohort
  leader: nia-1.0
  top tier: ['doxgpt-condensed', 'glass-5.6-max', 'nia-1.0', 'openevidence']
  Holm-adjusted p: doxgpt-condensed=0.5762, glass-5.6-max=0.3357, nia-1.0=1.0000, openevidence=0.5762

F1_weighted all 100 cases, clinical cohort
  leader: nia-1.0
  top tier: ['nia-1.0']
  Holm-adjusted p: doxgpt-condensed=0.0406, glass-5.6-max=0.0003, nia-1.0=1.0000, openevidence=0.0003

F1_weighted held-out 70 cases, clinical cohort
  leader: doxgpt-condensed
  top tier: ['doxgpt-condensed', 'nia-1.0']
  Holm-adjusted p: doxgpt-condensed=1.0000, glass-5.6-max=0.0003, nia-1.0=0.4800, openevidence=0.0003

F1_weighted, 24-model Main cohort
  leader: nia-1.0
  top tier: ['nia-1.0']