Extend sequential-equivalence coverage to Track B area and power objectives

Extend sequential-equivalence checking to the area-reduction and power-reduction objectives in the GateTruth Track B agentic-repair benchmark, which are currently not protected by that gate.

Background

GateTruth uses sequential equivalence checking for behavior-preserving Track B objectives to ensure that an agent's optimization does not alter the baseline design's functional behavior. The timing-closure objectives and the power-reduction objective are equivalence-gated, but the area-reduction objective is explicitly marked as not behavior-preserving and therefore receives only hidden-testbench simulation.

The paper identifies this as a disclosed verification gap: extending equivalence coverage to the remaining physical-optimization objectives would provide a stronger safeguard against functional regressions and make the Track B protocol more uniform.

References

We treat this as a genuine, disclosed gap in this task's design rather than paper it over: extending sec coverage to the area- and power-reduction objectives is future work (Section~\ref{sec:limitations}).

— GateTruth: Auditing the Rigor of RTL Design Benchmarks via Mutation Testing  (2608.12635 - Bhadra, 12 Aug 2026) in Section 6, Track B: Agentic Protocol; Section 10, Limitations and Threats to Validity

Settling this requires exhibiting a reference solution that passes both gates for each objective, which we have not authored; until then the honest statement is that no model produced a solution our gate accepts, not that no acceptable solution exists.

— GateTruth: Auditing the Rigor of RTL Design Benchmarks via Mutation Testing  (2608.12635 - Bhadra, 12 Aug 2026) in Section 8, Reference-Suite Baseline Results, Track B discussion