{
  "seed": 20261021,
  "n_train": 1600,
  "n_test": 800,
  "n_items": 30,
  "target": [0.5, 2],
  "arbitrary_cutoff": 1,
  "converged": true,
  "fit": {
    "M2": 453.990453724347,
    "df": 405,
    "p": 0.04651158820465,
    "RMSEA": 0.008697691256275,
    "RMSEA_5": 0.001214033421413,
    "RMSEA_95": 0.01275049122132,
    "SRMSR": 0.01814760902951
  },
  "mean_training_q3": -0.02906645562592,
  "adjusted_q3_threshold": 0.2,
  "enemy_pair_count": 0,
  "enemy_pairs_1based": [],
  "assembly": {
    "success": true,
    "status": 0,
    "stage1_floor": 5.594593382205,
    "floor_tolerance": 1e-07,
    "selected_items": [8, 10, 17, 20, 29, 30],
    "min_target_info_floor": 5.594593382205,
    "target_csem_profile": [0.422, 0.416, 0.411, 0.407, 0.404, 0.403, 0.401, 0.401, 0.401, 0.402, 0.404, 0.406, 0.409, 0.413, 0.417, 0.423]
  },
  "metrics": {
    "full": {
      "items": ["item_01", "item_02", "item_03", "item_04", "item_05", "item_06", "item_07", "item_08", "item_09", "item_10", "item_11", "item_12", "item_13", "item_14", "item_15", "item_16", "item_17", "item_18", "item_19", "item_20", "item_21", "item_22", "item_23", "item_24", "item_25", "item_26", "item_27", "item_28", "item_29", "item_30"],
      "indices_1based": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30],
      "domains": {
        "A": 10,
        "B": 10,
        "C": 10
      },
      "training_alpha": 0.9273307131382,
      "holdout_alpha": 0.9218463110563,
      "minimum_target_information": 10.871412186257,
      "rmse": 0.2421899147748,
      "bias": -0.01700683714112,
      "target_rmse": 0.2603299643177,
      "target_n": 234,
      "theta_correlation": 0.9702881083923,
      "sensitivity": 0.8134328358209,
      "specificity": 0.984984984985,
      "true_positive_n": 134,
      "true_negative_n": 666
    },
    "item_total": {
      "items": ["item_26", "item_17", "item_13", "item_15", "item_29", "item_22"],
      "indices_1based": [26, 17, 13, 15, 29, 22],
      "domains": {
        "A": 0,
        "B": 3,
        "C": 3
      },
      "training_alpha": 0.8166323091057,
      "holdout_alpha": 0.8046323638716,
      "minimum_target_information": 2.92417912185,
      "rmse": 0.4134182637385,
      "bias": 0.001018916068711,
      "target_rmse": 0.4062558265083,
      "target_n": 234,
      "theta_correlation": 0.9098141138809,
      "sensitivity": 0.7089552238806,
      "specificity": 0.966966966967,
      "true_positive_n": 134,
      "true_negative_n": 666
    },
    "content_matched": {
      "items": ["item_08", "item_07", "item_17", "item_13", "item_26", "item_29"],
      "indices_1based": [8, 7, 17, 13, 26, 29],
      "domains": {
        "A": 2,
        "B": 2,
        "C": 2
      },
      "training_alpha": 0.8079037381991,
      "holdout_alpha": 0.8012824030086,
      "minimum_target_information": 3.755006168996,
      "rmse": 0.4282691623136,
      "bias": -0.003163503300876,
      "target_rmse": 0.4171185772851,
      "target_n": 234,
      "theta_correlation": 0.9028622461588,
      "sensitivity": 0.6716417910448,
      "specificity": 0.966966966967,
      "true_positive_n": 134,
      "true_negative_n": 666
    },
    "constrained": {
      "items": ["item_08", "item_10", "item_17", "item_20", "item_29", "item_30"],
      "indices_1based": [8, 10, 17, 20, 29, 30],
      "domains": {
        "A": 2,
        "B": 2,
        "C": 2
      },
      "training_alpha": 0.7742220893142,
      "holdout_alpha": 0.7944026351222,
      "minimum_target_information": 5.594593382205,
      "rmse": 0.494793237316,
      "bias": -0.02185960108929,
      "target_rmse": 0.4562245952633,
      "target_n": 234,
      "theta_correlation": 0.8677099755684,
      "sensitivity": 0.6940298507463,
      "specificity": 0.978978978979,
      "true_positive_n": 134,
      "true_negative_n": 666
    }
  },
  "notes": ["All responses and item identities are synthetic; no clinical instrument or norms.", "Train and test are independent draws from one correctly specified unidimensional GRM population.", "No item text, readability assessment or DIF screening; zero grades disable the readability constraint.", "Content labels are metadata, not distinct latent dimensions. No induced local dependence.", "Single seed; no bootstrap or parameter uncertainty. Held-out scores use fixed training parameters and N(0,1) prior.", "Cutoff theta=1 is arbitrary; classification is against simulated true theta, not clinical status.", "Conditional information is fitted-model information, not empirical held-out accuracy."]
}