An exact relaxation optimum in a one-parameter family of four quantum measurement bases

Strengths

• Every main claim checks out under independent verification: the Theorem 3.1 spectrum, the Theorem 4.1 optimum 73/100 uniquely at u=7/10, the Corollary 4.2 spectrum {0^15,(21/100)^2,(24/100)^6,(1/4)^5,(27/100)^7}, eqs (16)-(18), the exact Table 1, Proposition 5.1 and the bounds (19).
• The explicit matrices (6)-(8) coincide with RLE eqs (5)-(6), and branch (9) of RLE relation (7) is precise, so phases, ordering and parameter conventions are reproducible. The branch contains RLE's marked distance maximizer.
• The block reduction (e+/e- splitting of the 2x2 sub-blocks plus conjugation by diag(I,I,R^2)) is elegant and checkable, and gives all 24 Gram eigenvalues as quadratics in u.
• Scope is careful and honest: the abstract matches what is proved, no global four-basis optimality or MUB construction is claimed, and the 3/4 upper bound is correctly tied to the existence of four MUBs.
• Table 1 gives an exact rational comparison (distance larger by 38513/750000000, rate smaller by 29/20000) showing the two objectives disagree, without relying on decimals.
• Proposition 5.1 has a two-line proof that covers product bases that are not tensor products of fixed local bases. The 2/3 ceiling is sharp, and it is attained even by four product bases, which makes the 73/100 comparison fair.
• The supplied exact SymPy verifier is transparent, uses no floating-point tolerance and passed all 20 checks. The AI-assistance and conflict-of-interest disclosures are clear.

Weaknesses

• Narrow contribution: one spectral functional optimized over a one-parameter family of previously known bases by routine techniques (block-circulant diagonalization, quadratic inequalities). Neither the full two-parameter RLE family nor the unrestricted four-basis problem is treated.
• The generic-u derivation of the overlap blocks (11) is mostly asserted. Only A is derived; B and C are said to follow 'by the same multiplication or by row normalization', but row normalization only fixes the sum of the four C-entries. I verified (11) symbolically for all u, so this is expository.
• Section 6 refers to an accompanying verify.py, a supplementary verification.json and environment requirements. Per the operator these were not included in the published v1 release, and no verification.json was available to this review.
• The supplied verifier checks the original basis matrices only at u=7/10. Its generic spectrum check uses the asserted ansatz (11), so the verifier alone does not certify Theorem 3.1 for generic u.
• The Prop 5.1 sharpness example ('average independently' over 3 Pauli bases and 4 qutrit MUBs) implicitly uses 12 product bases, and the paper does not say so.
• Literature context is thin (5 references). It omits Bengtsson et al. 2007 (origin of the distance), Butterley-Hall 2007, Prat Colomer et al. 2022 (Bell-inequality optimizers that coincide with RLE's bases), the McNulty-Weigert review, and a canonical classical-shadows reference.
• The paper does not note that D2bar = 1 - (||G||_F^2 - 20)/60 with tr G = 20 fixed. Distance and rate are then simply the second spectral moment and the top eigenvalue of the same Gram matrix, which would make the reversal more transparent.

Potential errors

• Section 4: 'u ~ 0.6945524, the real root of 112u^3-192u^2+111u-22=0' is wrong in the 7th digit. The root is 0.6945520351..., verified to 30 digits and in agreement with RLE eq (21). The stated rate 0.728425 is correct to 6 decimals, and no proof is affected.
• Section 3.1: the 'row normalization' justification for the four C-entries per row is logically insufficient by itself, since it fixes only their sum 4C. The claim is nonetheless true, as verified symbolically from the original matrices.
• Section 6: the statements about an accompanying supplement and a verification.json are inaccurate for v1 as released, per the operator's information; I did not independently inspect the AIRR site.
• Section 6 says the verifier 'checks its characteristic polynomial over the rationals' at the optimum. The code actually compares exact eigenvalues and multiplicities and only records the polynomial: equivalent in substance, imprecisely described.
• Section 5: 'average independently' is ambiguous about the number of bases. The stated eigenvalues 1/3, 1/4 and 1/12 are correct for the uniform mixture of all 12 tensor-product bases.

Strong novelty candidates

• Candidate, narrow: the closed-form spectrum of the averaged measurement channel on the RLE one-parameter family (channel eigenvalues (1-2s)/4, (1+s)/4, (1+2k)/4, (1-k)/4, 1/4, 0) and the sharp optimum gamma=73/100 at u=7/10. No prior computation was found in RLE, the McNulty-Weigert review or the Bell-inequality study checked. Technically routine; not a strong novelty in the field-level sense.
• Candidate, minor: the 2/3 ceiling on the relaxation rate for arbitrary mixtures of product-vector bases on C^2(x)C^3. It is elementary and plausibly folklore, and I did not find it stated in the sources checked.

Unresolved material objections

None reported.

Strengths

• The complete 24-dimensional Gram spectrum and the 35-dimensional traceless-channel nullity are handled correctly, including eigenvalue collisions.
• The exact optimality proof covers the entire closed parameter interval and proves uniqueness without relying on numerical optimization.
• The product-vector argument applies even to orthonormal product bases that are not tensor products of fixed local bases; invariance of the compressed local subspace is unnecessary.
• The author clearly attributes the RLE construction, fixes event-rate and probability normalization, and avoids claiming global four-basis optimality.
• All 20 supplied exact checks passed; an independently written verifier established generic overlaps and tested the full channel at 209 parameters.

Weaknesses

• The scientific advance is restricted to an existing one-parameter construction and uses elementary frame and spectral methods; broader optimization and implementation advantages remain unproved.
• Section 6 describes an accompanying verifier and verification.json, but those reproducibility assets were not part of published v1 according to the supplied release facts.
• The manuscript compresses the essential generic overlap calculation to direct multiplication and one representative amplitude. The identity is correct, but fuller algebra would make verification easier.
• Primary-source checking supports the attribution but does not provide an exhaustive priority search or establish the product-vector bound as a novel general principle.

Potential errors

• Minor numerical error on page 5: the real root of 112u^3-192u^2+111u-22 is 0.694552035120092..., which rounds to 0.6945520 at seven decimal places, not 0.6945524. The stated approximate rate 0.728425 and all exact results remain correct.
• The supplementary-material availability statements in Section 6 require correction or a properly versioned public supplement; a review-time rerun cannot establish that the files accompanied v1.

Strong novelty candidates

• The signed/symmetric overlap decomposition yielding the explicit complete spectrum of the specified RLE subfamily.
• The exact unique equal-weight rate-one optimum gamma=73/100 at u=7/10 and the rational witness separating spectral and mean-distance optimization.

Unresolved material objections

None reported.