{"id":"fd89293f-d490-405b-89e3-c587d14d93c8","arxiv_id":"2507.10688","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"In monitored free fermion circuits, the bipartite stabilizer mutual information shifts from logarithmic to constant scaling across the entanglement transition, while total magic remains extensive.","lead":"A numerical study of random free fermion circuits with measurements finds that a quantum 'magic' measure stays extensive through the measurement-induced transition, while a bipartite version of it switches from logarithmic to constant scaling and relaxes slowly in the critical phase. General readers may care because magic is a cost factor for quantum circuits, so tracking it in monitored dynamics informs quantum computing overhead.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The paper's own random-state formulas imply I1 and I2 are both approximately 2S_A; if this reduction holds for the simulated Gaussian states, the BSMI transition is inherited from entanglement and the central magic-delocalization claim is unsupported.","rationale":"The paper's central claim is that BSMI exhibits a magic delocalization transition concurrent with, but distinct from, the entanglement transition. The most direct threat to this claim is internal: the paper's own Appendix A analytic formulas, which it uses to set the sign convention in Eq. (4), imply that for states with a Gaussian Pauli spectrum the BSMI is approximately twice the entanglement entropy. In the critical phase S_A ~ log L, so I ~ 2 log L; in the area-law phase S_A ~ O(1), so I ~ O(1). The observed BSMI scaling is therefore exactly what the known entanglement transition produces, and no independent magic information is needed to explain it. The paper never checks this reduction on its simulated states; it only checks that BSMI scales similarly to entanglement entropy, which is the predicted consequence of the reduction rather than evidence against it. The concern is not that the numerics are wrong or that the algorithm is flawed; the perfect-sampling method is a real contribution. The concern is about interpretation: the headline 'magic delocalization transition' requires BSMI to be capturing more than entanglement, and that requirement is unverified. Because the central abstract claim is unsupported as stated, the reader's REJECT verdict is appropriate.","tokens_in":17203,"tokens_out":5370,"duration_ms":66519,"concrete_test":"For the stored covariance matrices used in Figs. 1 and 3, recompute S_A (second Rényi entropy) and the three SRE values M1(A), M1(B), M1(AB) from the same perfect sampler, and form δ(L) = I1(L) - 2S_A(L) for L = 32, 40, 48, 56, 64 at fixed |A|/L = 1/2. Repeat for β = 0.1 (critical) and β = 0.4 (area law). If δ stays small (say < 0.5 bit) while I1 rises by several bits over the same L range, the BSMI transition is a direct consequence of entanglement scaling; if δ grows with L, the reduction fails and the magic-delocalization claim has independent support.","verdict_should_be":"REJECT","load_bearing_attack":"The load-bearing step is the claim that BSMI measures nonlocal magic rather than just entanglement. Appendix A derives, for a Gaussian Pauli spectrum, M1 ≈ N + S and M2 ≈ N - S in the regime N - S ≫ 1. Applying this to a bipartite pure state gives M1(A) ≈ N_A + S_A, M1(B) ≈ N_B + S_B, and M1(AB) ≈ N, so Eq. (4) gives I1 ≈ 2S_A; with the sign convention in Eq. (4), I2 ≈ 2S_A as well. Thus the logarithmic-to-constant scaling of BSMI is exactly what the entanglement entropy already predicts. The paper uses App. A to justify the sign convention in Eq. (4), so it cannot dismiss this reduction as irrelevant. The actual steady states of the free-fermion circuits are not Haar-random, so the Gaussian-spectrum assumption might fail, but the paper presents no test of M1(A) ≈ N_A + S_A or of I1 = 2S_A on its own data. Without such a test, the central conclusion—that the structure of magic undergoes a delocalization transition distinct from the entanglement transition—is not established. The perfect-sampling implementation and the dynamics results remain valuable, but they do not rescue the interpretive claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the stabilizer Rényi entropy (SRE) and a bipartite stabilizer mutual information (BSMI) in 1+1D random free-fermion circuits with projective or weak measurements. Using a perfect sampling algorithm for Gaussian states, it reports that the total SRE remains extensive in both the critical and area-law phases, while the BSMI scales logarithmically in the critical phase and saturates to an O(1) constant in the area-law phase, mirroring the entanglement entropy. The paper interprets this as a magic delocalization transition concurrent with the entanglement transition. It also analyzes the dynamics of SRE, claiming a slower, universal relaxation (with saturation time ~ L) in the critical phase compared to generic random circuits.","tokens_in":17478,"tokens_out":6694,"duration_ms":72052,"significance":"The paper's strengths include a clean implementation of a perfect sampling algorithm for the SRE of fermionic Gaussian states, enabling computations up to L=64, and a concrete dynamical observation of slow magic saturation in free-fermion circuits (Sec. III B). If the central interpretive claim were established, the paper would offer an interesting perspective on non-local magic in monitored many-body systems. However, the central claim currently rests on the BSMI scaling, and the paper's own Appendix A implies that for states with a Gaussian Pauli spectrum, BSMI is approximately twice the entanglement entropy. Unless the authors demonstrate that this reduction fails for their steady states, the BSMI transition is not evidence for a distinct magic transition, but rather a restatement of the entanglement transition.","major_comments":[{"comment":"Appendix A derives, for a Gaussian Pauli spectrum, the asymptotic formulas M1 ≈ N + S and M2 ≈ N − S in the regime N − S ≫ 1. Applying these to a bipartite pure state and substituting into Eq. (4) gives I1 ≈ 2S_A and I2 ≈ 2S_A, where S_A is the (second Rényi) entanglement entropy of subsystem A. Since the sign convention in Eq. (4) is itself justified in the main text by the Appendix A results, the paper cannot dismiss this relation as irrelevant. The observed logarithmic-to-constant scaling of BSMI is then exactly the scaling of 2S_A, so the central claim that BSMI reveals a distinct magic delocalization transition is not established. The authors must test whether the relations M1(A) ≈ N_A + S_A and M2(A) ≈ N_A − S_A hold for the actual Gaussian steady states, or directly compare I1 and I2 with 2S_A on the same data. Without such a test, the BSMI transition is a restatement of the entanglement transition.","section":"II (Eq. (4)) and Appendix A"},{"comment":"The scaling claims—logarithmic vs. constant BSMI and the apparent transition between them—rest on data for L ≤ 64 with no reported sample counts, error bars, or quantitative fits. The perfect sampling algorithm is stochastic, and the distinction between log L and (log L)^2 scaling, which the authors themselves flag in footnote [37] as numerically difficult, is central to the interpretation if Iα ≈ 2S_A. The paper should provide error estimates, specify the number of samples, and perform fits with confidence intervals (e.g., for the effective scaling exponent as a function of measurement rate) to support the phase-transition claim.","section":"III A, Figs. 1, 3, 6"},{"comment":"The dynamical collapse ∆M(t)/L vs. t/L and the inferred saturation time tsat ∼ O(L log L) or O(L) are based on a few system sizes and no error bars. The paper itself states in Sec. III B that the late-time exponential decay cannot be numerically confirmed; this weakens the claim of a universal relaxation form. The authors should either provide the data with uncertainties and a clear goodness-of-fit measure for the collapse, or temper the universal-form claim accordingly.","section":"III B, Fig. 4"}],"minor_comments":[{"comment":"The caption of Fig. 8 repeats '(b) Non-unitary dynamics of I1 with projective measurements'; the last panel label should be (d).","section":"Fig. 8 caption"},{"comment":"The sign convention in Eq. (4) is presented as adopted for the studied regimes; it would be clearer to state explicitly that the non-negativity is empirical and not guaranteed by a subadditivity property.","section":"Sec. II, Eq. (4)"},{"comment":"In Appendix A, S denotes the log-purity (second Rényi entropy) of the reduced density matrix; the main text uses S_A for entanglement entropy. Please unify the notation to avoid confusion in the reduction Iα ≈ 2S_A.","section":"App. A"},{"comment":"The cross-ratio collapse in Fig. 3(d) is computed with the modified SRE in Eq. (30) that samples only fermionic operators in A and B; the text should state clearly that the collapse applies to this modified quantity, not to the standard BSMI defined in Eq. (4).","section":"Sec. III A, Eq. (30)"},{"comment":"The description of the perfect sampler would benefit from stating the number of samples used in each simulation and the statistical uncertainty of Mα estimates.","section":"Sec. II A"}],"recommendation":"major_revision","confidential_remarks":"The main technical concern—the BSMI ≈ 2S_A reduction—is serious and potentially undermines the central claim if it holds for the simulated states. I would ask the authors to test this relation explicitly on their data; if it holds, the claim of a magic delocalization transition should be withdrawn or substantially reframed. The paper also lacks statistical reporting throughout, which is unusual for a sampling-based calculation. Given the concurrent preprint [42] mentioned in the acknowledgment, the novelty of the magic-transition observation may be limited, but that is secondary to the correctness issue."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the SRE numerics are competent and the relaxation dynamics are a real finding, but the central claim—a magic delocalization transition distinct from the entanglement transition—does not survive contact with the paper's own Appendix A. Under the Gaussian Pauli spectrum assumption, M1 ≈ N + S and M2 ≈ N − S, so the BSMI defined in Eq. (4) becomes I1 ≈ 2S_A and I2 ≈ 2S_A. The log-to-constant scaling of BSMI is then exactly what the entanglement entropy already gives. The paper never tests this reduction on its simulated states; it merely asserts the BSMI captures non-local magic. That is a load-bearing gap.\n\nWhat is genuinely new: the perfect sampling algorithm is correctly implemented and lets them push SRE to L≈64; the observation that SRE remains extensive in both phases is clean; and the dynamics—SRE saturation time O(L) in the critical phase with a clean t/L collapse, versus O(log L) in Haar circuits—is an interesting, defensible result. The cross-ratio scaling of the two-interval SMI is also a nice numerical probe.\n\nSoft spots besides the main one: no sample counts or error bars anywhere, and the data collapse is by eye. The paper itself flags the sign ambiguity of SMI and the absence of the expected diffusive relaxation. Minor in comparison.\n\nI agree with the reader: the conceptual novelty is low once you see the reduction, and the paper does not provide the one plot or check that would settle it (e.g., I1 vs 2S_A on the same trajectory). The citation list is appropriate, and the acknowledgment of a competing preprint is honest. This is not a careless paper; it is an overinterpreted one.\n\nWho gets value: people working on magic in monitored circuits will want the dynamics numbers. The magic-transition framing should not be taken at face value. I'd send it to a serious referee because the question matters and the methods are sound enough to fix, but the referee should demand the reduction test. I would not cite the central claim.","headline":"The magic-transition claim is undercut by the paper's own Appendix A; the SRE dynamics results are the solid part.","tokens_in":18005,"tokens_out":3687,"would_cite":false,"duration_ms":42089,"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":"Monitored free-fermion circuits exhibit a magic delocalization transition that mirrors the entanglement transition.","keywords":["stabilizer Rényi entropy","magic transition","monitored free fermion circuits","measurement-induced entanglement transition","Gaussian states","bipartite stabilizer mutual information","non-stabilizerness","perfect sampling"],"falsifier":"Plot the bipartite stabilizer mutual information against the entanglement entropy $S_A$ for the same trajectories across the transition. If $I_1$ and $I_2$ collapse onto a single function of $S_A$ (in particular $I\\approx 2S_A$, as the paper's random-state formulas suggest), the magic transition is not independent; if the collapse fails, BSMI carries information beyond entanglement.","tokens_in":16961,"feed_emoji":"✨","tokens_out":11614,"duration_ms":124845,"temperature":0.7,"pith_summary":"Magic, the non-Cliffordness of a quantum state, is a resource as important as entanglement for quantum computation, but it is usually measured by a total count that hides spatial structure. This paper studies random free-fermion circuits with measurements in 1+1 dimensions, where a known measurement-induced transition separates a critical phase with logarithmic entanglement from an area-law phase. It claims that although the total stabilizer Rényi entropy stays extensive (volume-law) on both sides of that transition, the structure of magic undergoes its own delocalization transition: the bipartite stabilizer mutual information scales logarithmically with system size in the critical phase and saturates to a finite constant in the area-law phase, exactly mirroring the entanglement entropy. The paper also finds that magic approaches its steady state slowly, with a saturation time linear in system size in the critical phase, much slower than in generic random unitary circuits. If correct, this gives a magic-based order parameter for measurement-induced criticality and ties a quantum-computing resource to the entanglement transition.","feed_headline":"Magic undergoes a delocalization transition in monitored circuits","feed_subtitle":"Total magic stays extensive, while its nonlocal part turns from log L to a constant.","key_machinery":"The central object is the stabilizer Rényi entropy $M_\\alpha(\\rho)$, defined as the Rényi-$\\alpha$ entropy of the squared Pauli-string expectation values of the state; it vanishes on stabilizer states and measures non-Cliffordness. The paper's diagnostic for non-local magic is the bipartite stabilizer mutual information $I_\\alpha = \\pm(M_A + M_{\\bar A} - M_{A\\cup\\bar A})$, with the sign fixed by Eq. (4) so that $I_\\alpha\\ge0$ in the regimes studied; this combination cancels local magic and isolates magic supported jointly across a subsystem and its complement. The computation runs on a perfect-sampling algorithm that generates Majorana strings bit by bit using the chain rule of probability, with each marginal given by a determinant identity from the Gaussian covariance matrix, making SRE and BSMI computable in polynomial time for systems up to about one hundred sites. The circuits are random brickwork free-fermion unitary gates combined with local Z-basis projective or weak measurements, which preserve Gaussianity and realize the known entanglement phase transition.","core_discovery":"The central discovery is a magic localization-delocalization transition that occurs at the same measurement rate as the entanglement transition in monitored free-fermion circuits. Using the stabilizer Rényi entropy (SRE) as the magic measure and a perfect-sampling algorithm that draws Majorana strings from the Gaussian state's covariance matrix, the authors compute both the total SRE and the bipartite stabilizer mutual information (BSMI) for circuits with projective and weak local Z measurements. The total SRE remains extensive in both phases, but the BSMI grows as $\\log L$ in the critical phase and saturates to a finite O(1) constant in the area-law phase; for two disjoint intervals in the critical phase, the BSMI collapses onto a function of the cross ratio with power-law exponents roughly $0.75$ for $I_1$ and $0.93$ for $I_2$. Dynamically, the SRE deviation from its steady state collapses as a function of $t/L$, giving a saturation time $O(L)$ in the critical phase and $O(L\\log L)$ under purely unitary free-fermion evolution, in contrast to the $O(\\log L)$ saturation seen in generic random unitary circuits.","pith_inferences":["The paper's own random-state formulas imply $M_1\\approx N+S$ and $M_2\\approx (N-S)/(\\alpha-1)$, so both $I_1$ and $I_2$ reduce to about $2S_A$; if that reduction holds for the simulated Gaussian states, the BSMI scaling would be inherited from the entanglement entropy rather than an independent magic phenomenon.","A direct test would normalize BSMI by the entanglement entropy $S_A$ across the transition: a constant ratio would mean BSMI is a derived quantity, whereas a ratio that changes at the critical point would establish magic delocalization as an independent order parameter.","Because magic is the resource that makes a state costly for stabilizer-circuit simulation, a magic delocalization transition suggests that the classical simulation cost of monitored free-fermion states changes qualitatively at the entanglement transition even though the states remain Gaussian and efficiently representable.","The cross-ratio collapse of BSMI suggests that nonlocal magic in the critical phase may be captured by the same conformal field theory that describes the entanglement criticality; identifying the CFT quantity that yields the observed exponents would connect magic to universal data."],"forward_implications":["Total stabilizer Rényi entropy is extensive in both phases, so the measurement-induced transition is invisible to total magic and requires a nonlocal diagnostic such as BSMI.","The bipartite stabilizer mutual information acts as an order parameter for the concurrent magic transition: $\\log L$ scaling in the critical phase and a finite constant in the area-law phase, with the transition point matching the entanglement transition.","In the critical phase, nonlocal magic encodes universal data: the BSMI of two disjoint intervals collapses onto a function of the cross ratio, with exponents roughly $0.75$ for $I_1$ and $0.93$ for $I_2$.","Magic relaxation is parametrically slower in free-fermion circuits than in generic random unitary circuits, with saturation time $O(L)$ in the critical phase and $O(L\\log L)$ under purely unitary evolution.","Projective and weak measurements qualitatively reproduce the same steady-state and dynamic scaling, so the magic delocalization transition is robust to the measurement protocol."],"supporting_citations":[{"why":"Defines the stabilizer Rényi entropy used throughout as the magic measure.","marker":"[8]"},{"why":"Supplies the perfect-sampling algorithm that makes SRE computable for fermionic Gaussian states.","marker":"[24]"},{"why":"Also cited for the perfect-sampling method for SRE in related magic-transition settings.","marker":"[25]"},{"why":"Establishes the critical phase and its conformal and cross-ratio properties in nonunitary free-fermion dynamics.","marker":"[13]"},{"why":"Provides the free-fermion monitored entanglement transition and the scaling behavior discussed in the paper's footnote on logarithmic corrections.","marker":"[14]"},{"why":"Gives the fast $O(\\log L)$ magic saturation baseline in generic random unitary circuits that the paper's slower relaxation results are measured against.","marker":"[38]"}],"fun_headline_variants":["Magic delocalizes at entanglement transition in fermion circuits","Stabilizer mutual info reveals magic delocalization in free fermions","Monitored fermions: magic stays extensive, its structure delocalizes","Log L magic mutual info saturates at measurement-induced transition"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the sign convention in Eq. (4) makes the bipartite stabilizer mutual information a meaningful independent measure of non-local magic; if the random-state relation $I\\approx 2S_A$ holds for the simulated states, the observed log-versus-constant scaling would simply restate the entanglement transition.","fun_headline_variants_meta":{"raw":{"variants":["Magic delocalizes at entanglement transition in fermion circuits","Stabilizer mutual info reveals magic delocalization in free fermions","Monitored fermions: magic stays extensive, its structure delocalizes","Log L magic mutual info saturates at measurement-induced transition"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000705,"raw_usage":{"total_tokens":3194,"prompt_tokens":979,"completion_tokens":2215,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":595,"completion_tokens_details":{"reasoning_tokens":2141}},"tokens_in":595,"tokens_out":2215,"duration_ms":18659,"temperature":1.0,"reasoning_tokens":2141,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T17:28:32.374212+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Plot the bipartite stabilizer mutual information against the entanglement entropy $S_A$ for the same trajectories across the transition. If $I_1$ and $I_2$ collapse onto a single function of $S_A$ (in particular $I\\approx 2S_A$, as the paper's random-state formulas suggest), the magic transition is not independent; if the collapse fails, BSMI carries information beyond entanglement.","supporting_citations":[],"review_version":1}