Create benchmark_spec.json

benchmark_spec.json

@@ -0,0 +1,159 @@
+{
+  "repo_name": "clinical-five-node-mof-cascade-boundary-v1.0",
+  "repo_path": "ClarusC64/clinical-five-node-mof-cascade-boundary-v1.0",
+  "version": "v1.0",
+  "conceptual_upgrade": "Closed-Loop Control Geometry",
+  "concept_summary": "v1.0 evaluates whether a model can select and execute a complete closed-loop control sequence across a five-node multi-organ-failure cascade under live feedback, adaptive timing, and durable recovery constraints.",
+  "domain": "clinical",
+  "structure": "five-node-cascade",
+  "variables": [
+    "metabolic_stress",
+    "perfusion_deficit",
+    "renal_strain",
+    "hepatic_drift",
+    "buffer_capacity"
+  ],
+  "failure_type": "mof_boundary",
+  "label_column": "label_mof_boundary",
+  "primary_metric": "recall_correct_control_sequence_selection",
+  "secondary_metric": "false_effective_control_rate",
+  "default_label_rule": "label = 1 if stabilization_success = 1 AND trajectory_shift < -0.10 AND intervention_alignment_score >= 0.60 AND control_sequence_alignment_score >= 0.60 AND recovery_consistency_score >= 0.60",
+  "mid_strength_label_rule": "label = 1 if stabilization_success = 1 AND trajectory_shift < -0.10 AND control_sequence_alignment_score >= 0.60",
+  "relaxed_label_rule": "label = 1 if stabilization_success = 1",
+  "trajectory_signals": [
+    "drift_gradient",
+    "drift_velocity",
+    "drift_acceleration",
+    "trajectory_shift"
+  ],
+  "boundary_signals": [
+    "boundary_distance",
+    "secondary_boundary_distance",
+    "boundary_competition_ratio"
+  ],
+  "uncertainty_signals": [
+    "boundary_uncertainty",
+    "trajectory_uncertainty",
+    "regime_confidence",
+    "transition_uncertainty",
+    "intervention_uncertainty",
+    "controller_confidence",
+    "feedback_noise_ratio"
+  ],
+  "regime_transition_signals": [
+    "regime_transition_score",
+    "transition_direction",
+    "regime_separation_margin",
+    "transition_velocity"
+  ],
+  "intervention_signals": [
+    "intervention_leverage_score",
+    "intervention_alignment_score",
+    "rescue_window_width",
+    "pathway_divergence_margin",
+    "intervention_competition_ratio",
+    "primary_intervention_path",
+    "secondary_intervention_path",
+    "pathway_switch_velocity",
+    "minimal_intervention_path"
+  ],
+  "control_signals": [
+    "control_sequence_alignment_score",
+    "control_horizon",
+    "feedback_response_score",
+    "intervention_timing_score",
+    "adaptation_latency",
+    "control_stability_margin",
+    "sequence_divergence_margin",
+    "controller_confidence",
+    "recovery_consistency_score",
+    "control_recalibration_count",
+    "terminal_pathway_state",
+    "feedback_noise_ratio",
+    "controller_oscillation_score",
+    "rollback_trigger_count"
+  ],
+  "recovery_signals": [
+    "recovery_distance",
+    "recovery_gradient",
+    "return_feasibility"
+  ],
+  "perturbation_signals": [
+    "perturbation_radius",
+    "collapse_trigger"
+  ],
+  "delta_signals": [
+    "delta_metabolic_stress",
+    "delta_perfusion_deficit",
+    "delta_renal_strain",
+    "delta_hepatic_drift",
+    "delta_buffer_capacity"
+  ],
+  "outcome_fields": [
+    "stabilization_success",
+    "terminal_pathway_state",
+    "label_mof_boundary"
+  ],
+  "withheld_in_tester": [
+    "stabilization_success"
+  ],
+  "thresholds": {
+    "control_alignment_threshold": 0.6,
+    "intervention_alignment_threshold": 0.6,
+    "recovery_consistency_threshold": 0.6,
+    "high_uncertainty_threshold": 0.25,
+    "narrow_window_threshold": 0.3,
+    "oscillation_threshold": 0.25,
+    "stable_margin_threshold": 0.6,
+    "trajectory_shift_threshold": -0.1
+  },
+  "scorer_entrypoint": "scorer.py",
+  "scorer_mode": "binary_classification_with_closed_loop_control_diagnostics",
+  "reference_metrics": [
+    "accuracy",
+    "precision",
+    "recall",
+    "f1",
+    "confusion_matrix",
+    "recall_correct_control_sequence_selection",
+    "false_effective_control_rate",
+    "primary_intervention_path_accuracy",
+    "secondary_intervention_path_accuracy",
+    "control_sequence_alignment_accuracy",
+    "control_horizon_error",
+    "feedback_response_accuracy",
+    "intervention_timing_accuracy",
+    "high_uncertainty_control_miss_rate",
+    "narrow_window_control_miss_rate",
+    "adaptation_latency_error",
+    "control_stability_error",
+    "recovery_consistency_error",
+    "recalibration_overuse_rate",
+    "controller_oscillation_misread_rate",
+    "terminal_pathway_state_accuracy"
+  ],
+  "evaluation_diagnostics": [
+    "primary_intervention_path_accuracy",
+    "secondary_intervention_path_accuracy",
+    "control_sequence_alignment_accuracy",
+    "control_horizon_error",
+    "feedback_response_accuracy",
+    "intervention_timing_accuracy",
+    "high_uncertainty_control_miss_rate",
+    "narrow_window_control_miss_rate",
+    "adaptation_latency_error",
+    "control_stability_error",
+    "recovery_consistency_error",
+    "recalibration_overuse_rate",
+    "controller_oscillation_misread_rate",
+    "terminal_pathway_state_accuracy"
+  ],
+  "files": [
+    "data/train.csv",
+    "data/tester.csv",
+    "scorer.py",
+    "benchmark_spec.json",
+    "dataset_schema.json",
+    "README.md"
+  ]
+}