{
  "artifact": "SZLHOLDINGS/szl-formulas surrogate v1",
  "role": "21-class formula identifier from call-trace observables \u2014 kernel remains ground truth",
  "generator": {
    "script": "scripts/forge.py",
    "seed": 20260721,
    "kernel_version": "0.1.0",
    "kernel_labelled": true,
    "kernel_registry_count": 21,
    "kernel_replay_audited_samples": 126,
    "label_source": "each sample is produced by calling the REAL REGISTRY[name] evaluator; label == producing formula"
  },
  "data": {
    "rows": 18900,
    "n_classes": 21,
    "classes": [
      "bekenstein_cascade",
      "bohr_complementarity_floor",
      "css_ingress_verify",
      "dsse_envelope",
      "fisher_rao_distance",
      "gleason_quantum_lambda",
      "hoeffding_tail",
      "khipu_merkle_root",
      "kitaev_surface_correct",
      "kochen_specker_18vector_witness",
      "lambda_aggregate",
      "lambda_bounded",
      "lambda_homogeneous",
      "madhava_series",
      "pac_bayes_mcallester",
      "pinsker_kl_bound",
      "reed_solomon_singleton",
      "reidemeister_invariant",
      "schur_concave_lambda_two_axis",
      "shor_codeword_distance",
      "two_witness_ks18_soundness"
    ],
    "per_class_samples": 900,
    "class_counts": {
      "bekenstein_cascade": 900,
      "bohr_complementarity_floor": 900,
      "css_ingress_verify": 900,
      "dsse_envelope": 900,
      "fisher_rao_distance": 900,
      "gleason_quantum_lambda": 900,
      "hoeffding_tail": 900,
      "khipu_merkle_root": 900,
      "kitaev_surface_correct": 900,
      "kochen_specker_18vector_witness": 900,
      "lambda_aggregate": 900,
      "lambda_bounded": 900,
      "lambda_homogeneous": 900,
      "madhava_series": 900,
      "pac_bayes_mcallester": 900,
      "pinsker_kl_bound": 900,
      "reed_solomon_singleton": 900,
      "reidemeister_invariant": 900,
      "schur_concave_lambda_two_axis": 900,
      "shor_codeword_distance": 900,
      "two_witness_ks18_soundness": 900
    },
    "split": "80/20 stratified",
    "features": [
      "arity",
      "argshape_n",
      "argshape_depth",
      "argshape_is_str",
      "argshape_is_bytes",
      "argshape_is_dict",
      "argshape_is_list",
      "argshape_is_bool",
      "argnum_count",
      "argnum_mean",
      "argnum_std",
      "argnum_min",
      "argnum_max",
      "argnum_frac_zero",
      "outshape_n",
      "outshape_depth",
      "outshape_is_str",
      "outshape_is_bytes",
      "outshape_is_dict",
      "outshape_is_list",
      "outshape_is_bool",
      "outnum_count",
      "outnum_mean",
      "outnum_std",
      "outnum_min",
      "outnum_max",
      "outnum_frac_zero",
      "out_is_bool",
      "out_is_int",
      "out_is_float",
      "out_is_bytes",
      "out_is_dict",
      "out_is_list",
      "out_scalar_logabs"
    ],
    "feature_policy": "trace observables ONLY: argument-shape + kernel-output shape/type/aggregate stats; no formula name leaked into features"
  },
  "model": {
    "type": "sklearn.HistGradientBoostingClassifier",
    "params": {
      "max_iter": 300,
      "early_stopping": true,
      "random_state": 20260721
    },
    "file": "model.joblib",
    "sha256": "94cca70dc950fa957c0931ee80240c4369cab0ff8446ffbfedb483c3e2907698"
  },
  "metrics_MEASURED": {
    "test_accuracy_21_class": 0.9709,
    "mean_per_class_recall": 0.9709,
    "per_class_recall": {
      "bekenstein_cascade": 1.0,
      "bohr_complementarity_floor": 0.7778,
      "css_ingress_verify": 1.0,
      "dsse_envelope": 1.0,
      "fisher_rao_distance": 0.8889,
      "gleason_quantum_lambda": 1.0,
      "hoeffding_tail": 1.0,
      "khipu_merkle_root": 1.0,
      "kitaev_surface_correct": 1.0,
      "kochen_specker_18vector_witness": 1.0,
      "lambda_aggregate": 0.9889,
      "lambda_bounded": 1.0,
      "lambda_homogeneous": 1.0,
      "madhava_series": 1.0,
      "pac_bayes_mcallester": 1.0,
      "pinsker_kl_bound": 0.8833,
      "reed_solomon_singleton": 1.0,
      "reidemeister_invariant": 1.0,
      "schur_concave_lambda_two_axis": 0.85,
      "shor_codeword_distance": 1.0,
      "two_witness_ks18_soundness": 1.0
    },
    "blind_spots": {
      "policy": "classes with held-out recall < 0.90 reported honestly",
      "classes": {
        "bohr_complementarity_floor": 0.7778,
        "fisher_rao_distance": 0.8889,
        "pinsker_kl_bound": 0.8833,
        "schur_concave_lambda_two_axis": 0.85
      },
      "statement": "formulas that share an argument+output OBSERVABLE signature (e.g. several bool-returning axiom checks over similar inputs) are confusable from trace shape alone; the kernel's evaluators remain authoritative for identity"
    }
  },
  "environment": {
    "python": "3.12.12",
    "sklearn": "1.9.0",
    "numpy": "2.5.1",
    "host": "replit 2-vCPU container",
    "wall_seconds": 4.3
  },
  "honesty": "Every number above is MEASURED by this run. The surrogate identifies formulas from trace shape; it never replaces the kernel's real evaluators. \u039b untouched = Conjecture 1.",
  "trained_at_utc": "2026-07-21T02:50:06Z"
}