baseline/example_predictions.csv

scenario_id,predicted_outcome_mode,predicted_first_drift_step,predicted_mechanism_family,predicted_repair_outcome,predicted_residual_constraint_failure,confidence
RDOD2-025,visible_failure,3,premature_constraint_closure,unrepaired,Preserve multiple plausible causes until discriminating evidence appears.,0.91
RDOD2-026,no_drift,,none,not_applicable,none,0.94
RDOD2-027,recovered_drift,3,causal_scope_collapse,partial_repair,Do not infer causal direction from correlation.,0.89
RDOD2-028,recovered_drift,3,evaluation_scope_collapse,full_repair,none,0.9
RDOD2-029,visible_failure,3,multi_objective_collapse,unrepaired,Do not use average improvement to erase concentrated disadvantage.,0.91
RDOD2-030,no_drift,,none,not_applicable,none,0.92
RDOD2-031,visible_failure,3,evaluation_scope_collapse,failed_repair,Require evidence across subgroups shift calibration and operational conditions.,0.9
RDOD2-032,recovered_drift,3,proxy_substitution,full_repair,none,0.9
RDOD2-033,latent_drift,4,causal_scope_collapse,failed_repair,Do not attribute change to the most visible intervention alone.,0.87
RDOD2-034,recovered_drift,3,proxy_substitution,partial_repair,Separate signature from informed understanding.,0.88
RDOD2-035,no_drift,,none,not_applicable,none,0.94
RDOD2-036,no_drift,,none,not_applicable,none,0.94