Could architectural changes enforcing dimensional analysis or symmetry constraints reduce the gap on research-scale problems?

Your question is subtle and touches something important, but the preprints here don't directly address it—they're about quantum optimization and annealing, not dimensional analysis or symmetry as *architectural* design principles for closing research-scale gaps.

That said, I can sketch what *does* emerge from the archive:

**Symmetry constraints *are* being used to reduce complexity.** In quantum phase estimation [AG-2024.03-1561], researchers exploit symmetries in chemical Hamiltonians to tighten spectral bounds—achieving smaller scaling factors than naive worst-case estimates. This is a concrete win: exploiting symmetry structure *reduced the resource cost* of an algorithm on realistic problems. The logic: dimensional analysis (here, the spectral norm) becomes looser when you ignore problem structure; tighter when you enforce it.

Architectural embedding leverages connectivity constraints. Papers [AG-2024.05-2370], [AG-2024.04-1625], and [AG-2024.03-1560] tackle the reverse problem: mapping a problem's geometry onto hardware graphs. They show that *matching* the problem's constraints to hardware topology—using parity embeddings and optimized chain strengths—improves scaling. This is implicitly enforcing dimensional/symmetry alignment.

Divide-and-conquer through separators [AG-2024.05-1606] is another angle: partitioning a graph to reduce inter-device communication cost hinges on identifying bottlenecks (separators), which is fundamentally a structural/symmetry argument.

But none of these papers frame the question as: "Could we *design* algorithms or hardware around dimensional analysis from the ground up to close research-scale gaps?"—the research here is reactive, not prescriptive architecture.