Retaining Repair Hypotheses for Cost-Aware Execution

A repair-to-execution interface weights compatible finite guard repairs using diagnostic and ordinary-execution observations to select optimal routes or tests. Testing costs for workload-valued scenarios decreased by 3.33 percent compared to random selection when using an AI-weighted belief. This method reduces no-test costs by 2.14 percent over single language-model-generated repairs.