{"id":"8c9cfd7f-a8c9-4228-8565-88298a0c90a8","arxiv_id":"2607.15739","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Cat codes break the global GKP magic-non-Gaussianity equivalence, but a sector-adapted SU(d) asymptotic cat code locally aligns Wigner-logarithmic-negativity contours with magic-measure contours.","lead":"Cat codes do not inherit the exact, global link between non-Gaussianity and magic that GKP codes have. The paper shows that by choosing a specially rotated cat-code basis, the two resources can still be locally aligned, and argues this is easier to reach experimentally than with approximate GKP states.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Local-alignment construction assumes without proof that the top eigenvector of A_ε lies inside the sign sector; for d≥3 this can fail, making the ACC alignment undefined.","rationale":"The reader's weakest_assumption correctly identifies the unproven sector-eigenvector assumption. My stress-test converges on the same point: the proof of Lemma 3 (Appendix C) establishes only that within a fixed sign sector the magic measure is a linear functional of the A_ε eigenvalue probabilities; it does not establish that the top eigenvector vmax actually carries the sign pattern ε, nor that the sign sector is nonempty. The qubit case is verified explicitly, but the qutrit example is not checked, and no general argument is supplied. Since the SU(d) ACC construction (Eq. (31)) is the paper's main constructive contribution, this is the most load-bearing gap. Other potential issues, such as the heuristic nature of the contour-deviation benchmark and the absence of code/data, are secondary and do not affect the mathematical claim. The asymptotic WLN formula (Theorem 1) is well supported by the high-frequency averaging argument in Appendix A, and the d=2 global-incompatibility theorem is sound. Thus the appropriate verdict remains CONDITIONAL: the paper should either prove the sector-admissibility condition, state and prove a criterion for which sign patterns yield vmax ∈ S_ε, or explicitly restrict the alignment claim to d=2 with a conjecture for higher d. My read does not change the reader's verdict.","tokens_in":18736,"tokens_out":6794,"duration_ms":55898,"concrete_test":"Perform a numerical scan for d=3: for each of the 2^9 sign patterns ε (or at least the pattern used in Eq. (28)), construct A_ε = Σ ε_lm O_lm, diagonalize it, and compute the sector signature s_lm = sign Tr(O_lm |vmax⟩⟨vmax|) for the largest-eigenvalue eigenvector vmax. Also check whether any s_lm = 0. If for some ε with nonempty S_ε one finds s ≠ ε, then vmax ∉ S_ε and the ACC construction for that sector is invalid. If all tested patterns satisfy s = ε, the concern is empirically refuted for d=3; a proof would still be needed for general d.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central d-dimensional result — the SU(d) ACC local alignment of Sec. III B — rests on identifying the magic-measure 'local maximizer' vmax (top eigenvector of the sector signature operator A_ε, Eq. (20)) with a coherent component, and on the claim that within S_ε the phase orbits around vmax organize the level sets. This requires vmax ∈ S_ε and S_ε nonempty. Appendix C proves only the conditional statement: for a state already in S_ε, the magic measure is a weighted sum of A_ε eigenvalues (Lemma 3, Eq. C8), so M depends on amplitudes. It does not show that the unconstrained maximizer of that quadratic form has the sign pattern ε. Indeed the proof's phrase 'Since |ψ⟩ also lies in the ε sector' assumes the very point at issue for vmax. For d=2, the example is explicitly verified, but for general d no argument is given that the distinguished sign pattern is realizable or that vmax is not on a sign boundary. If vmax lies outside S_ε, the sector's local maximum occurs on a boundary, the U(1)^{d−1} phase rotations around vmax leave the sector, and the aligned level-set picture described after Eq. (31) is not realized. The qutrit example in Eq. (28) does not check the signature of vmax either. This is a load-bearing gap in the paper's central claim that local alignment is 'constructive' in arbitrary dimension.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies whether the magic–non-Gaussianity correspondence established for GKP codes survives in cat codes. It computes the asymptotic Wigner logarithmic negativity [Eq. (10)] for non-degenerate d-peaked cat states, shows that its level sets are governed by coherent-state amplitudes and have a U(1)^{d−1} phase-rotation symmetry, and contrasts this with a magic measure defined through a phase-operator basis, whose level sets are organized by sign sectors and sector eigenbases. For the even/odd qubit cat code the two global foliations are incompatible (Theorem 2). The paper then constructs an \"SU(d) asymptotic cat code\" [Eq. (31)] whose sector eigenbasis is identified with the coherent-state basis, and claims a constructive local alignment between WLN and magic-measure level sets. A numerical benchmark compares the circularity of finite-cat and approximate-GKP WLN contours.","tokens_in":19090,"tokens_out":16484,"duration_ms":137238,"significance":"The asymptotic WLN formula is clean; the averaging lemma is sound and the non-overlap argument is justified. The qubit incompatibility theorem is correct, and the observation that any magic–NG relation is encoding-dependent is a useful conceptual point. If the d-dimensional local-alignment construction were fully justified, the paper would provide a genuine geometry-based route beyond GKP. However, two load-bearing points need repair: the sector-eigenvector assumption and the mismatch between L1-type WLN level sets and quadratic magic-measure level sets for d≥3. With these fixed, or with the claims appropriately scoped, the paper would be a solid contribution.","major_comments":[{"comment":"The construction assumes vmax ∈ S_epsilon. Lemma 3 only proves a conditional statement: if |ψ⟩ is already in S_epsilon, then M depends on the amplitudes. It does not show that the largest-eigenvalue eigenvector of A_epsilon has the sign pattern epsilon, nor that S_epsilon is nonempty. If vmax lies outside S_epsilon, the unconstrained maximum is not in the sector, the sector's local maximum is on a boundary, and the phase orbits around vmax leave the sector, so the ACC alignment is undefined. The qubit example verifies the property; the qutrit example (Eq. (28)) does not check the signature of vmax, and no general argument is given. Please prove this property or restrict the construction to sectors for which it holds.","section":"Sec. III.B, Lemma 1/3 (Appendix C), Eq. (31)"},{"comment":"For d≥3 the claimed 'level-set alignment' is not established. After the identification (30), WLN∞ = log2[1 + (4/π)Σ_{i<j}|a_i a_j|] and dM = λ_max − Σ_j(λ_max−λ_j)|a_j|² are different functions of the amplitudes. Near vmax the WLN has a term linear in Σ_{j>0}|a_j|, while M is a smooth quadratic form. The level sets are therefore different surfaces; the two functions share only the U(1)^{d−1} phase-rotation symmetry (torus orbits). For d=2 a single transverse radius makes this distinction invisible, but for general d the abstract's and Sec. III.B's 'local level-set alignment' overstates what is proven. Please either prove a genuine level-set matching under the ACC or explicitly formulate the result as phase-orbit compatibility.","section":"Sec. III.B, Eq. (10), Appendix C Eq. (C8)"}],"minor_comments":[{"comment":"The abscissa labels are garbled ('j,j2') and the dual scale (|α| and Δ/σ) is confusing; please redraw with a clear dual axis and legend.","section":"Fig. 4"},{"comment":"U_epsilon is a unitary rather than strictly SU(d); specify the global-phase convention that makes the code SU(d).","section":"Eq. (31)"},{"comment":"There are at most 2^{d²} sign sectors, and many may be empty; write 'at most' rather than O(2^{d²}) unless counting nonempty sectors.","section":"Sec. III.A, after Eq. (19)"},{"comment":"The factor d in d·M arises from the normalization in Eq. (C4); state this explicitly to avoid confusion.","section":"Appendix C, Eq. (C8)"},{"comment":"The lemma should not refer to vmax as the local maximum unless vmax ∈ S_epsilon is added as a hypothesis; as written, Lemma 1 is valid for arbitrary states in the sector, and the vmax claim is a separate unproven assertion.","section":"Lemma 1 statement"}],"recommendation":"major_revision","confidential_remarks":"The core issue is not a fatal error: the qubit results and the asymptotic WLN are correct, and the sector-eigenvector problem may be fixable by restricting the construction to 'good' sectors or by choosing the sector after verifying the signature of vmax. However, the d≥3 level-set language needs to be revised downward to phase-orbit compatibility unless new arguments are supplied. The manuscript's main conceptual message (encoding-dependence and local phase-orbit alignment) survives in a weakened form, and the technical gaps are identifiable and localizable, so major revision is appropriate."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is worth a serious look for two things: the asymptotic WLN formula for d-peaked cat states in Theorem 1, and the argument that cat codes do not inherit the global GKP magic–NG correspondence. Theorem 2, the qubit incompatibility result, is correct as stated. The WLN derivation is clean: the high-frequency averaging lemma is standard, and the non-overlap of Gaussian envelopes in the large-separation limit is handled carefully. I credit that.\n\nThe soft spot is exactly where the reader put it. The local SU(d) alignment in Sec. III B is presented as constructive, but it depends on an unproved assumption: that the top eigenvector vmax of the sector signature operator A_epsilon lies inside the sign sector S_epsilon, and that S_epsilon is nonempty. Appendix C only proves a conditional statement: if a state is already in the sector, then the magic measure depends on amplitudes. It does not show that the maximizer of that quadratic form has the sign pattern epsilon. For d=2 the example is checked explicitly, but the qutrit example does not check the signature of vmax. So the general-d local-alignment claim is not established. This is a load-bearing gap, not cosmetic, because the ACC basis is defined by vmax. The paper should either prove the sector-eigenvector assumption, confine the alignment claim to d=2, or state it as a conjecture for d>2.\n\nThe finite-parameter benchmark is plausible but weaker. The 'geometric error' is a custom metric, there is no code or data file, and the comparison with approximate GKP is not a general resource comparison. The claim that cat codes reach the tolerance at moderate |alpha| is not independently checkable from the text alone. Still, this is a secondary point.\n\nThe citation pattern looks appropriate: the GKP magic–NG correspondences are cited, and the new content is clearly distinguished. The paper is honest about what is asymptotic and what is heuristic.\n\nWho should read it? People working on CV resource theories, especially magic–NG correspondences. They will get a clear picture of cat-code geometry and a useful caution that the correspondence is encoding-dependent. The paper deserves a referee, but the referee should insist that the local-alignment claim be either proven or scaled back.","headline":"Useful asymptotic WLN result and a correct qubit-level incompatibility claim, but the general-d local alignment rests on an unproved sector-eigenvector assumption.","tokens_in":623,"tokens_out":1005,"would_cite":true,"duration_ms":25714,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Cat codes do not inherit the global, value-preserving GKP magic–non-Gaussianity equivalence, but a sector-adapted SU(d) asymptotic cat code aligns the local level-set geometry of Wigner logarithmic negativity with that of the phase-operator","keywords":["cat codes","Wigner logarithmic negativity","magic resource","non-Gaussianity","GKP encoding","asymptotic cat code","phase-operator basis","level-set alignment"],"falsifier":"Take a concrete dimension (e.g. d=3) and the sign pattern of Eq. (28), diagonalize Aε, and test whether vmax satisfies sgn[Tr(Ol,m|vmax⟩⟨vmax|)] = εl,m for every l,m; a single sign mismatch would show the claimed sector has no state at the supposed maximizer. Then compute exact finite-|α| WLN contours and magic-measure contours under the SU(d) ACC on a grid; if the contour-radius deviations do not vanish as |α|→∞ in the target sector, the local alignment claim fails.","tokens_in":18581,"feed_emoji":"🐈","tokens_out":6703,"duration_ms":54694,"temperature":0.7,"pith_summary":"This paper asks whether the exact magic–non-Gaussianity correspondence discovered for GKP codes survives in cat codes, and it answers both yes and no. It proves that in the large-separation limit a d-peaked cat state has Wigner logarithmic negativity WLN∞ = log2[1 + (4/π)Σ|cicj|], so the WLN level sets are organized by coefficient amplitudes and invariant under relative phase rotations. The phase-operator magic measure, by contrast, is organized locally by sign sectors and sector-dependent eigenbases, so the two global level-set structures cannot be matched. The constructive part is that both resources share a local U(1)^{d-1} phase-rotation invariance; by choosing the cat-code basis from the magic measure's sector eigenbasis through an SU(d) transformation, the WLN contours and magic-measure contours align locally in the chosen sign sector. The paper also benchmarks physical finite-amplitude cat codes against approximate GKP states and finds that moderate cat amplitudes reach the same contour-deviation tolerance as much larger GKP scale separations.","feed_headline":"Cat codes break the global magic–non-Gaussianity link","feed_subtitle":"Local contours of Wigner negativity and magic align under a chosen SU(d) cat code, at accessible amplitudes.","key_machinery":"The paper's central objects are (i) the asymptotic WLN formula of Theorem 1, obtained by high-frequency averaging of pairwise interference fringes in Eq. (9); (ii) the sign-sector decomposition of the magic measure, generated by the Hermitian signature operator Aε = Σ εl,m Ol,m, whose largest-eigenvalue eigenvector is the local magic maximizer and whose eigenbasis carries a U(1)^{d-1} phase-rotation invariance (Lemma 1); and (iii) the SU(d) asymptotic cat code |μ⟩(∞)ε,L = Σi (Ũε)μi|αi⟩, which aligns these two local geometries by mapping the sector eigenbasis onto the coherent-component axes. The WLN level-set structure supplies the rigid circular geometry; the signature-operator eigenbasis s","core_discovery":"Under the non-degenerate large-separation limit, the paper claims, the Wigner function of a d-peaked cat state decomposes into separated Gaussian peaks and interference fringes, each fringe's absolute value averaging to 2/π. This yields the closed-form asymptotic WLN of Theorem 1, which depends only on the amplitudes |ci| and is invariant under all relative phases: the equal-WLN surfaces are circular level sets in coefficient space. The phase-operator magic measure, however, is organized by sign sectors; inside a sector its value depends only on amplitudes in the eigenbasis of the signature operator Aε, with a local U(1)^{d-1} phase invariance. The paper shows these two global landscape stru","pith_inferences":["A natural extension is to define a sector-by-sector alignment map for cat codes, treating the magic–NG relation as a piecewise local correspondence patched across sign sectors rather than a single global function.","Because the alignment depends only on amplitude and phase-rotation invariance, the same SU(d) construction should apply to other multi-component coherent-state superpositions, not just displaced equal-amplitude cats.","The contour-deviation benchmark suggests a practical experimental probe: measuring the circularity of equal-WLN contours in generated cat states would directly test whether the local magic–NG alignment is present, without demanding GKP-scale squeezing.","The two-sector alignment seen in the qubit example may be exceptional; an interesting open question is how many sign sectors can be simultaneously aligned in higher dimensions d."],"forward_implications":["No universal, encoding-independent magic–non-Gaussianity relation exists: the form of any such relation depends on the bosonic code, since cat codes and GKP codes give incompatible global level-set structures.","For any dimension d and any sign sector, one can construct a sector-adapted SU(d) cat code whose asymptotic WLN level sets locally coincide with the magic measure's level sets in that sector.","The local alignment is geometric rather than value-preserving, so high non-Gaussianity does not correspond to high magic in the aligned sector; the WLN minimum is placed at the magic maximum.","In finite-parameter physical realizations, the SU(2) cat code reaches a 1% contour-deviation tolerance at |α|≈2.04, while the approximate GKP family considered here needs Δ/σ≈20.7 under the same metric, suggesting the cat route is more accessible."],"fun_headline_variants":["Cat codes: global magic–NG split, local alignment","Cat codes lose global magic link, align locally","Non-Gaussianity vs magic: cat codes part ways globally","Cat-code resources: global clash, local geometric alignment"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The alignment construction assumes that, for the chosen sign sector, the largest-eigenvalue eigenvector vmax of the signature operator Aε lies inside that sector and is its local magic maximizer; the paper does not prove this for arbitrary d, and if vmax falls outside Sε the local level-set alignment is not realized.","fun_headline_variants_meta":{"raw":{"variants":["Cat codes: global magic–NG split, local alignment","Cat codes lose global magic link, align locally","Non-Gaussianity vs magic: cat codes part ways globally","Cat-code resources: global clash, local geometric alignment"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001597,"raw_usage":{"total_tokens":6211,"prompt_tokens":763,"completion_tokens":5448,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":507,"completion_tokens_details":{"reasoning_tokens":5383}},"tokens_in":507,"tokens_out":5448,"duration_ms":34517,"temperature":1.0,"reasoning_tokens":5383,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T22:28:04.138701+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a concrete dimension (e.g. d=3) and the sign pattern of Eq. (28), diagonalize Aε, and test whether vmax satisfies sgn[Tr(Ol,m|vmax⟩⟨vmax|)] = εl,m for every l,m; a single sign mismatch would show the claimed sector has no state at the supposed maximizer. Then compute exact finite-|α| WLN contours and magic-measure contours under the SU(d) ACC on a grid; if the contour-radius deviations do not vanish as |α|→∞ in the target sector, the local alignment claim fails.","supporting_citations":[],"review_version":1}