Measure false-rejection rates of RTL benchmark testbenches

Measure the false-rejection rate of RTL benchmark testbenches by evaluating independently correct alternative implementations or formally specified behaviors, thereby determining whether mutation-certified testbenches wrongly reject valid designs.

Background

GateTruth's mutation score measures sensitivity to injected faults, but a high kill rate does not establish that a testbench correctly accepts all valid implementations. The paper distinguishes this missing specificity or false-rejection analysis from mutation sensitivity.

For RTLLM v2.0, the authors lack independently correct alternatives or formal specifications needed to perform the measurement. They explicitly identify this as unresolved future work.