Per-level five-number summary (min / Q1 / median / Q3 / max)

- intermediate_health_literacy (n=80)
  - factual_attribution: min=0.0000 Q1=0.4926 median=0.5926 Q3=0.7605 max=0.9412
  - completeness: min=0.1000 Q1=0.9252 median=1.0000 Q3=1.0000 max=1.0000
  - info_quality: min=0.0000 Q1=0.4507 median=0.5714 Q3=0.7157 max=0.9412

- proficient_health_literacy (n=80)
  - factual_attribution: min=0.0000 Q1=0.4901 median=0.7207 Q3=0.8213 max=0.9600
  - completeness: min=0.0000 Q1=0.8750 median=1.0000 Q3=1.0000 max=1.0000
  - info_quality: min=0.0000 Q1=0.4565 median=0.6682 Q3=0.7792 max=0.9286

- low_health_literacy (n=80)
  - factual_attribution: min=0.0000 Q1=0.3656 median=0.5263 Q3=0.6667 max=1.0000
  - completeness: min=0.2400 Q1=0.9365 median=1.0000 Q3=1.0000 max=1.0000
  - info_quality: min=0.0000 Q1=0.3433 median=0.5000 Q3=0.6376 max=1.0000

Suggested 1D ordinal thresholds (low / intermediate / proficient)

- factual_attribution: t1=0.5590 t2=0.6132 dir=increasing (balanced_acc=0.483)
  - low: x < 0.5590
  - intermediate: 0.5590 ≤ x < 0.6132
  - proficient: x ≥ 0.6132
- completeness: t1=0.9583 t2=0.8775 dir=decreasing (balanced_acc=0.400)
  - low: x ≥ 0.9583
  - intermediate: 0.8775 ≤ x < 0.9583
  - proficient: x < 0.8775
- info_quality: t1=0.5119 t2=0.5731 dir=increasing (balanced_acc=0.479)
  - low: x < 0.5119
  - intermediate: 0.5119 ≤ x < 0.5731
  - proficient: x ≥ 0.5731

Decision tree rules (multivariate, shallow):
|--- source_coverage <= 0.68
|   |--- conciseness <= 0.09
|   |   |--- class: 2
|   |--- conciseness >  0.09
|   |   |--- class: 0
|--- source_coverage >  0.68
|   |--- conciseness <= 0.60
|   |   |--- class: 2
|   |--- conciseness >  0.60
|   |   |--- class: 1