Bayesian Matroid-Union Bounds for Quantum List Discrimination: Support Congestion, Process Compression, and Exact Adaptive-Parallel Phases

Strengths

• Lemma 3.2 and Theorem 4.1, pp.4-5: substituting the Rado rank into Edmonds union and writing B=C disjoint-union F leaves the nonnegative term (ell-1)|F|. Combining rho_i<=P_A with the per-outcome budget and s_i<=1 therefore gives exactly the stated union-rank inequalities, including mixed supports and soft rewards.
• Proposition 4.3, p.6: summation by parts over sorted priors makes each weighted prefix deficit nonnegative. Equality is consequently equivalent to saturation only at strict prior drops; zero priors and ties cause no hidden extra equality conditions.
• Lemma 5.1, Theorem 5.2 and Corollary 5.3, pp.6-8: restricting inverse square roots to supp T correctly produces a POVM and normalized effective states. Common congruence cannot raise any support-sum rank. The positive core argument remains valid for subnormalized cores, and the separate tail suprema are explicitly distinguished.
• Theorems 6.1-6.2, pp.8-10: common input dephasing permits a label-independent flag decomposition, including reference systems. Conditional adaptive memory contains no information beyond the flagged transcript. Laminar cells give q+1 parallel classes and feedback gives at most 2^q leaves; the breadth-first and prefix-following constructions attain both bounds for ell=2^s.
• Theorems 7.1-7.2, pp.10-12: the complete-error-basis twirl gives S=D I tensor rho_R and a constant PGM overlap, with a matching dual. The Weyl fibre sizes and terminal support dimensions correctly give the stated SER0/PAR0/PARBell laws under their explicit restrictions.
• Theorem 9.2 and Proposition 10.1, pp.14-16: the root-frame effects sum to identity and annihilate every omitted simplex vertex. The qubit circumcenter has radius 1/sqrt(3), positive barycentric weights and matching exclusion primal/dual cost. These examples establish both attainment and genuine insufficiency of full union rank.

Weaknesses

• Sections 11 and 13, pp.16-18, describe exact rational/algebraic checks of complete-unitary-basis frame spectra and a fixed-probe certificate. The supplied verify_multitime_ueb.py constructs neither frame matrices nor POVM effects/dual slacks; it uses floating square roots, tolerances and equality of formula values. The successful replay is much weaker evidence than that description.
• Section 11 describes 240 small subspace instances, but verify_list_matroid_union.py uses only three support fixtures and counts their subset/list-budget combinations. The number 240 is correct as a case count, not as the number of independently varied subspace families.
• Section 7 and related work: Feng-Duan-Ji already derive and attain the approximate dense-coding formula (Eq. (17), https://arxiv.org/pdf/quant-ph/0604149). The paper acknowledges this; the process-list translation and restricted wiring consequences should carry the novelty claim, rather than the spectrum formula itself.
• Section 8: the projector obstruction of Stratton-Hsieh-Skrzypczyk can be stronger than the union density threshold, as the paper acknowledges. In the qubit example the largest eigenvalue of sum rho_i is (15+sqrt(153))/10>2, already excluding a perfect two-list POVM. The new value calculation is useful, but the impossibility is not a new feasibility criterion.
• Sections 5 and 13: full-rank perturbations can trivialize support bounds; selecting useful positive cores remains an extra optimization. The 81/100 cap is not shown attainable, and no optimal algorithm for general priors, noise or unrestricted process architectures is proved.
• Priority and packaging: the one-label predecessor [3] was not retrieved, so incremental overlap remains unverified. The supplied batch contains the replay scripts but not the LaTeX/bibliography package advertised on p.18; a manuscript rebuild was therefore not performed. No stored or author-provided report was treated as independent evidence.

Potential errors

• Section 7, p.12: the D=4 rank-two fixture is described as an intermediate entanglement spectrum, although its nonzero Schmidt spectrum is flat (1/2,1/2). It demonstrates nonsaturation of the coarse list cap at intermediate Schmidt rank, not a nonflat-spectrum effect. The numerical value (2+sqrt(2))/4 itself is correct.
• Section 6, p.9: the insertion proof for q+1 laminar cells needs the inclusion-compatible order to put containing sets before contained sets; an arbitrary insertion can meet several existing cells. A parent-first order exists and makes the proof work, so this is a minor clarification, not a counterexample to the phase law.

Strong novelty candidates

• Theorems 4.1-4.3 and Corollary 5.3: inclusion vectors for arbitrary-prior mixed-state soft-list decisions lie in a single exact union-matroid independence polytope, with a prefix equality audit and a positive-core tail extension. The contribution is the operational synthesis, not new matroid theory.
• Theorem 6.2: the exact all-depth list/feedback phase against arbitrary parallel entanglement and adaptive quantum memories is a clear, fully bounded application. It remains a classical-query mechanism embedded in channels, as explicitly stated.
• The juxtaposition of an all-size simplex attainment family, the exact hemisphere optimum and the intermediate-rank list-cap failure gives concrete boundaries between support combinatorics and POVM geometry; priority for the package remains provisional.

Unresolved material objections

None reported.

Strengths

• Theorem 4.1 correctly combines the operator inequality rho_i <= P_A, the per-outcome reward budget, and singleton bounds to obtain exactly the independence polytope of the ell-fold Rado-matroid union.
• Lemma 5.1 preserves all tester probabilities under Moore-Penrose compression, while Theorem 5.2 correctly uses a common linear map to obtain a tester-independent but explicitly coarser support cap.
• The regular-simplex root-frame POVM and the three-qubit primal-dual exclusion certificate rigorously show both exact attainment and failure of support sufficiency within transparent finite examples.
• The laminar-channel theorem distinguishes parallel cells from adaptive decision-tree leaves and proves matching constructions against the stated entangled-parallel and quantum-memory strategy classes.

Weaknesses

• The adaptive part of Theorem 6.1 relies on a concise genie-flag induction; writing the conditional memory state and its label-independence invariant explicitly would make the reduction robust to instrument and transcript conventions.
• The D=4 rank-two list optimum in Equation (36) is sketched through sector pinching and a covariant dual; the full normalization across four shift sectors is not displayed in the manuscript even though the fixture checks the value.
• Several potentially new components are direct combinations of classical matroid union, support domination, and standard tester compression, so the contribution is substantial synthesis rather than a new foundational discrimination formalism.

Potential errors

• Corollary 8.1 refers to Theorem 4.2, although the relevant arbitrary-prior statement is Corollary 4.2; this appears to be a cross-reference error rather than a mathematical flaw.
• The phrase 'complete input-dependent family' in the abstract could be read broadly, while Theorem 6.2 concerns one specially constructed laminar family with uniform priors and power-of-two list sizes.

Strong novelty candidates

• Embedding the entire inclusion-probability vector into the ell-fold Rado-matroid independence polytope, with arbitrary priors and soft reward budgets, is the clearest general candidate contribution.
• The exact all-depth laminar channel phase against arbitrary entangled parallel inputs and adaptive quantum memories is a concrete architecture-sensitive novelty candidate.
• The paired simplex-attainment and hemisphere-obstruction examples give a sharp, reusable account of where the support obstruction is exact and where quantum geometry remains decisive.

Unresolved material objections

None reported.