metadata/evaluation_normal.json

{
  "subset_name": "evaluation_normal",
  "source": "official",
  "curators": [],
  "difficulty": "normal",
  "problem_count": 200,
  "true_count": 100,
  "false_count": 100,
  "selection_method": "human_ai",
  "derived_from": "stage1_final_evaluation",
  "notes": "Stage 1 final evaluation subset."
}