Complete Rank-Two Born-Prediction Rate–Distortion on Gr_C(2,4): All-Field Matrix–Bingham Rigidity and a Unique Coexistence Transition

Strengths

• Pages 3-4, Lemma 3.1/Theorem 3.2: the Dirichlet-sphere series has nonnegative coefficients, and strictness in h_2 forces a simplex vertex. The inverse pair-sum transform gives precisely the balanced spectrum, including degenerate spectra.
• Pages 4-6, Theorem 5.1: conditional least squares permits arbitrary standard-Borel memories; the balanced radius stays in the full capped density-matrix body. The reverse covariant law has the required Haar marginal and posterior mean.
• Pages 6-8, Lemma 6.1/Theorem 6.2: the exact S_n identity leaves one negative coefficient and a positive tail. F prime equals H, so the fold and unique positive contact genuinely follow from an all-degree argument, not numerical roots.
• Page 9, Theorem 7.1: constant channel divergence establishes the worst-source upper bound; independent Haar inputs, the chain rule and convexity supply the unrestricted joint-block lower bound.
• Page 10, Corollary 8.1 and calibration: the normalizer gives the pre-log four and constant 4 log(2/D)-4-log 12. The rank-one query second moment gives the factor 20 and endpoint error 1/80 consistently.

Weaknesses

• Pages 10-11: the supplied JSON hashes are d8ace26d...069f and 2f8f6a6f...7475, whereas the PDF prints 99dd5ad2...3f3cd and 71a0d7a3...4f1b1. All files match PUBLIC-CODE-MANIFEST.json; that does not reconcile the different release-freeze claims.
• The reproducibility commands use absent work/tenth_paper paths. I ran unchanged scripts from explicit copied batch locations; this is successful script execution, not successful replay of the documented source-archive layout.
• The recent Grassmann information-geometry PDF could not be retrieved past OpenReview verification. The unpublished companion manuscripts were not available in this batch, and I did not establish exhaustive novelty or their independent correctness.
• The original numerical verifier retains a statement that it does not claim contact uniqueness; the separate phase verifier and analytic proof do address uniqueness. The diagnostic descriptions should be made consistent.

Potential errors

• Page 7, Theorem 6.2: at lambda=0, phi_0(b)=0 for every b in [0,1], contradicting the displayed maximizer set {0}. Restrict the uniqueness classification to positive lambda and state the degenerate endpoint separately; R(1/4)=0 is unaffected.
• make_frontier_figure.py, k_prime: the small-k code uses k/15+k^3/1050+k^5/113400, whereas exact expansion is k/15+k^3/6300-k^5/189000+O(k^7). The plotted positive branch starts above kappa_c, so this erroneous small-field branch is not used there.

Strong novelty candidates

• Theorem 3.2: unique finite-dimensional, all-field external-spectrum maximization on the full traceless Frobenius sphere for Gr_C(2,4).
• Theorems 5.1 and 6.2 together: an unrestricted exact RDF with a proved unique coexistence face and no later radial exchange on this rank-two orbit.

Unresolved material objections

None reported.

Strengths

• Lemma 3.1 and Theorem 3.2 give a transparent global spectrum comparison: every series coefficient is nonnegative, h_k(z) is bounded by the fixed simplex sum, and equality in the positive k=2 term forces a single pair-sum square, hence the balanced 2+2 spectrum.
• Theorem 5.1 does not restrict reproductions to the source orbit: conditional least squares identifies the full posterior-mean body, and the scalar dual maximum is attained by an explicit covariant reverse channel whose posterior mean is I/4+bS_Q.
• Theorem 6.2 is stronger than a local cumulant diagnosis. The displayed coefficient formula has one negative nonzero term followed by positive terms, which makes the normalized fold numerator strictly increasing through one zero and forces a unique positive coexistence contact.
• Theorem 7.1 correctly uses Haar input for the converse and constant-divergence covariant product channels for achievability; the chain rule, one-letter RDF bound, convexity, and monotonicity yield nR(D) against arbitrary joint memories.
• Fresh execution reproduced complete_radial_phase.json exactly and returned the stated fold/contact values; the balanced verifier also passed its exact simplex, moment, transform, HCIZ, quadrature, and contact checks.

Weaknesses

• The priority discussion relies partly on two unpublished companion manuscripts and a limited neighboring bibliography. Exact-title and phrase searches did not establish exhaustiveness, so the claimed combination should be treated as plausible rather than independently certified priority.
• The Fenchel-Moreau coverage argument in Theorem 5.1 compresses existence, closure, and subgradient details for arbitrary standard-Borel memories. The route is standard and internally consistent, but a precise cited theorem with verified hypotheses would improve auditability.
• Theorem 7.1 applies the one-letter RDF to memory (W,P^{t-1}) while retaining a report depending only on W. This is valid under the earlier arbitrary-report formulation, but the proof should say so explicitly because Equation (10) is written using posterior means.
• The release instructions name workspace paths and a source archive that are not included in this batch, so only the supplied replay subset—not the claimed complete source snapshot—could be checked.

Potential errors

• The reproducibility section prints 99dd5ad2100f16ddf0cd57fd8f0178ec88635cad27f4e4eb4fe1d4035013f3cd for balanced_grassmann_verification.json, but the supplied file hashes to d8ace26d... and my fresh output to fa29f425...; the mismatch is real and not a successful source-hash verification.
• The same section prints 71a0d7a3e0f4aff4376f7a9aa50e630c7a67c298c71a32cbd5aea7d421f4f1b1 for complete_radial_phase.json, while both supplied and freshly generated files hash to 2f8f6a6f54f45bd1ad0e4a108e01e23ecd2ea558c965c0409360515879167475.
• The balanced verifier uses seeded floating linear algebra and root finding; its exact checks passed, but its output hash changed because the maximum residual and last digits of the contact diagnostic changed across the current numerical stack. A frozen byte hash is therefore not portable evidence for that output.

Strong novelty candidates

• The coefficientwise all-field proof that the balanced (2,2) spectrum uniquely maximizes the complete Gr_C(2,4) matrix-Bingham normalizer at every fixed Frobenius radius.
• The exact unrestricted two-piece RDF with one discontinuous coexistence face followed by a smooth matrix-Bingham branch, together with the no-reentrance theorem.
• The equality between the Haar-source RDF and minimum worst-source capacity, including exact n-letter tensorization against arbitrary noncausal joint classical memory.

Unresolved material objections

None reported.