{"id":"0fa7900a-cf14-4c08-9dcc-fd4e2c7e8bcf","arxiv_id":"2411.08857","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Classical mutual information between phase-space subsystems of a kicked top grows at about half the Lyapunov exponent in chaotic regions and slowly in regular regions, mirroring known signatures of bipartite entanglement.","lead":"The authors define a classical measure of nonseparability, mutual information between two pieces of angular momentum in a kicked top, and show it grows like quantum entanglement does: linear at roughly half the Lyapunov rate in chaotic regions, slower in regular regions. The work supports the view that chaos in quantum systems has a purely classical statistical counterpart.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reported mutual information is computed between scalar x-components (X1,X2), not between the full phase-space distributions defined as the classical analogue of entanglement; this projection is an unverified bottleneck for the central analogy.","rationale":"The reader's weakest assumption identifies exactly the point I find most load-bearing: the numerics measure a scalar projection, not the full phase-space nonseparability that the Introduction and Summary claim. This is an internal inconsistency, not merely a disagreement with consensus. The data processing inequality makes the direction of the gap precise, but the size of the gap is uncontrolled without a direct calculation. A second concern, the claimed growth rate ~0.5 lambda, is also real but less fundamental: the four values of lambda in Fig. 7 are all within 0.974--0.978, so the comparison supports linear growth but cannot meaningfully determine the proportionality constant. The scalar-projection issue is more basic because it questions what quantity is actually being connected to entanglement entropy; quantum linear entropy is basis-independent and distribution-level, while scalar MI is neither. I do not think this warrants rejection: the qualitative linear-versus-log growth and the phase-space dependence of equilibrium MI are visually supported, and the missing full-distribution calculation is straightforward with the provided code. The reader's CONDITIONAL verdict is the appropriate level; my reading does not move it.","tokens_in":14995,"tokens_out":4387,"duration_ms":44782,"concrete_test":"Using the released code, recompute the mutual information for the same initial ensembles and dynamics as in Figs. 7 and 9, but estimate the joint density over the full 4D angular variables (theta1, phi1, theta2, phi2), e.g., by a 64x64x64x64 histogram on the sphere product or a k-NN estimator in the 4D space. Compare the growth slope to the Lyapunov exponent and the equilibrium map to the classical phase portrait. If the full-distribution MI reproduces the ~0.5 lambda linear growth and the same regular/chaotic ordering, the scalar projection is validated; if the slope changes materially or the equilibrium pattern shifts, the central analogy is not established. As a secondary check, repeat with Y- and Z-components to test coordinate invariance.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The Introduction defines the classical analogue of entanglement as nonseparability of the full phase-space densities rho1(theta1,phi1) and rho2(theta2,phi2), and the Summary restates the result as \"mutual information between the marginal phase space densities of individual subsystems.\" However, all numerical results in Sec. 4 estimate I12 between the scalar variables X1 = J1x/j and X2 = J2x/j (Fig. 6 caption and surrounding text), not between the full marginal densities. This is a strict information bottleneck: for deterministic functions f and g, I(f(A);g(B)) <= I(A;B), so the reported quantity is at best a lower bound on the distribution-level mutual information that the paper claims to study. The paper provides no argument that this bound is tight, nor that its growth rate or equilibrium values track the full quantity. The quantum benchmark, linear entropy, is invariant under local unitary rotations of each spin, whereas scalar mutual information depends on the chosen coordinate frame; no invariance or robustness argument is supplied. The choice of the x-component is therefore an unverified modeling assumption directly beneath the paper's headline claim of \"completely analogous signatures\" of chaos.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a classical analogue of bipartite entanglement based on the mutual information between phase-space distributions of two subsystems of the kicked top. The authors define nonseparability of the joint classical distribution and compute mutual information between variables on the two sides of a bipartition, finding: (i) logarithmic growth of mutual information for regular dynamics and linear growth at a rate approximately half the positive Lyapunov exponent for chaotic dynamics; (ii) equilibration time scaling as O(√j) for regular and O(ln j) for chaotic dynamics; and (iii) equilibrium mutual information maps that visually resemble classical phase portraits and quantum entanglement entropy maps. The paper argues that these are completely analogous signatures to those of quantum entanglement in the kicked top.","tokens_in":15221,"tokens_out":2658,"duration_ms":26399,"significance":"If the central claims hold, the paper provides a concrete classical statistical analogue of entanglement that reproduces three well-known chaos signatures, potentially clarifying the classical-quantum correspondence in chaotic systems. The paper is well-structured, includes reproducible numerical procedures (codes and data are provided), and the qualitative features in Figs. 6 and 9 are visually clear. However, the quantitative claims—especially the proportionality between mutual-information growth rate and Lyapunov exponent—rest on a narrow set of numerical experiments and on an unverified projection from full phase-space distributions to scalar x-components. The conceptual contribution is valuable, but the current support for the quantitative part is weak.","major_comments":[{"comment":"The mutual information that is actually computed is I(X1;X2) for the scalar variables X1 = J1x/j and X2 = J2x/j, not the mutual information between the full marginal phase-space densities ρ1(θ1,φ1) and ρ2(θ2,φ2) that the Introduction defines as the classical analogue of entanglement and that the Summary claims to report. For deterministic functions f and g, I(f(A);g(B)) ≤ I(A;B), so the scalar quantity is only a lower bound on the distribution-level mutual information. The paper provides no argument that this lower bound is tight or that its growth rate and equilibrium values track the full quantity. This is load-bearing because the central claim of 'completely analogous signatures' concerns the distribution-level measure.","section":"Sec. 4, Fig. 6"},{"comment":"The claim that the growth rate of I12 is proportional to the Lyapunov exponent, specifically ~0.5×λ, is weakly supported. In Fig. 7, the four chaotic cases at κ=6.0 have Lyapunov exponents 0.974, 0.976, 0.978, and 0.976—nearly identical values. Consequently, the data demonstrate a single growth rate in an almost-uniform chaotic phase space, but cannot establish proportionality to λ. The same issue affects Fig. 8, where the four exponents at κ=2.5 (0.139–0.167) vary by only about 20%. To support the proportionality claim, the authors should present results across a range of λ values (e.g., different κ or different phase-space regions with greater Lyapunov-exponent contrast) and provide quantitative fits with uncertainties.","section":"Sec. 4, Fig. 7"},{"comment":"The equilibrium mutual information I12_eq is estimated by averaging I12 over the fixed window 400≤T≤500 for both κ=0.5 and κ=2.5. The paper does not demonstrate that this window corresponds to a true equilibrium for all initial conditions, nor does it provide error bars or tests of stationarity. Since the equilibration time scales differently for regular and chaotic dynamics (O(√j) vs. O(ln j)), using the same window for both cases may bias the comparison. The qualitative resemblance to Figs. 3 and 4 is suggestive, but the quantitative claim that equilibrium values are systematically higher in chaotic regions requires evidence of convergence.","section":"Sec. 4, Fig. 9"},{"comment":"The initial angular spread of subsystem 1 is set to sinθ0 Δθ Δφ = 1/4, while subsystem 2 uses the quantum-analogous spread 1/j. This choice is arbitrary and is a free parameter in the model. The paper does not test how the reported growth rates, equilibration times, or equilibrium maps depend on this spread. Since the analogy to a spin-1/2 subsystem motivates ∥J1∥=1/2 but not a specific phase-space area, a robustness study varying this spread (while keeping j fixed) is needed to establish that the observed signatures are not artifacts of this particular choice.","section":"Sec. 4, Fig. 5"}],"minor_comments":[{"comment":"In both equation groups, the three recursion relations are all labeled with J'_1x; the second and third lines should be J'_1y and J'_1z, respectively.","section":"Eqs. (9) and (14)"},{"comment":"The paper consistently spells 'Lyapunov' as 'Lyuapunov' (e.g., in the Abstract, Sec. 3, and Figs. 7–8). This should be corrected.","section":"Throughout"},{"comment":"The mutual information is first defined for random variables X1 and X2 generically, then later used specifically for X1 = J1x/j and X2 = J2x/j. The notation is confusing because the Introduction describes full marginal densities ρ1(θ1,φ1) and ρ2(θ2,φ2). Please clarify explicitly that the numerical implementation uses the scalar projection and explain why this is a valid proxy, or revise the definition to match the computed quantity.","section":"Sec. 1, Introduction"},{"comment":"The plots show growth rates but no confidence intervals or error bars for the slope estimates. Given that the proportionality factor ~0.5 is a quantitative claim, providing standard errors or bootstrap intervals would strengthen the result.","section":"Figs. 7 and 8"},{"comment":"The summary states that mutual information grows logarithmically for regular dynamics, but Fig. 6(a) shows substantial oscillations superimposed on the growth. The text should acknowledge that the logarithmic behavior is an overall envelope, not a pointwise fit.","section":"Sec. 5, Summary"}],"recommendation":"major_revision","confidential_remarks":"The paper has a clearly stated and potentially useful idea, and the qualitative figures are attractive. The main risk is that the scalar projection (X1, X2) may not faithfully represent the distribution-level mutual information that the paper claims to study; this is not just a presentation issue but affects the headline claim. The quantitative proportionality to the Lyapunov exponent also needs a stronger test. I would encourage the authors to either compute the full phase-space mutual information (or provide a rigorous argument for the scalar projection) and to expand the Lyapunov-exponent range in their numerical experiments. If those issues are resolved, the paper could be a solid contribution to the classical-quantum correspondence literature."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Thanks for the report. I read the paper and the stress-test note. My take: the paper's qualitative claim is probably right, and the authors have done the numerical work carefully enough to deserve a referee. But the stress-test note lands on a real gap, and the Lyapunov proportionality is thinner than the text implies.\n\nWhat's new: the general idea of classical mutual information as an entanglement analogue appears in refs. 13, 18, 19, 45, and the paper says so. The contribution here is the specific bipartition of the kicked top and the demonstration that the three standard entanglement signatures of chaos—linear versus logarithmic growth, equilibration time scaling with j, and higher equilibrium values in chaotic regions—all show up in this classical quantity. That is worth having, and the figures make it visually persuasive. The paper is also honest: it lists limitations and future directions, and the code is on GitHub with a detailed supporting information section.\n\nThe soft spots are real but not fatal. The main one is the projection. The introduction defines nonseparability via the full marginal densities ρ1(θ1,φ1) and ρ2(θ2,φ2), but every numerical result in Sec. 4 estimates I12 between the scalar variables X1 = J1x/j and X2 = J2x/j. Since I(f(A);g(B)) ≤ I(A;B), the reported quantity is a lower bound on the claimed distribution-level mutual information, and no argument shows the bound is tight or that the x-component is the right frame. The quantum benchmark, linear entropy, is invariant under local rotations of each spin; scalar mutual information is not. The stress-test note is correct that this is an unverified modeling assumption directly beneath the headline claim. It may well be that the scalar captures the qualitative behavior, but the paper needs to either compute the full quantity or justify the projection.\n\nThe second soft spot is the Lyapunov proportionality. The paper states the growth rate is ~0.5×λ, but Fig. 7 uses four chaotic cases at κ=6.0 whose Lyapunov exponents are all between 0.974 and 0.978. That's essentially one data point, with no error bars. Fig. 8 at κ=2.5 shows larger deviations, but still only four cases. The qualitative linear-vs-log distinction is solid; the factor of 0.5 is not established. The equilibration time scaling, O(√j) vs O(ln j), is shown in Fig. 6 but not fitted statistically; I'd like to see log-log fits or at least more j values.\n\nNone of this invalidates the qualitative picture. The paper is a legitimate contribution and the authors are clearly thinking carefully. I recommend sending it to peer review. A good referee should ask for a robustness check on the projection and a more careful treatment of the Lyapunov claim, but this is not a desk reject.","headline":"A clean numerical demonstration that classical mutual information in the kicked top tracks entanglement signatures of chaos, but the scalar projection used to compute it is an unverified bottleneck and the Lyapunov proportionality claim is weak.","tokens_in":15757,"tokens_out":2831,"would_cite":false,"duration_ms":24418,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["81Q50","37D45","94A17"],"pacs":["05.45.Mt","03.67.-a"],"model":"deepseek-v4-flash","headline":"Classical mutual information mimics entanglement's chaos signature","keywords":["classical analogue of entanglement","mutual information","kicked top","quantum chaos signatures","phase-space distributions","Lyapunov exponent","nonseparability","entanglement-chaos correspondence"],"falsifier":"Compute $I_{12}$ between other component pairs ($Y_1$ and $Y_2$, or $Z_1$ and $Z_2$) for the same initial conditions and compare growth rates and equilibrium maps with the $\\sim 0.5\\lambda$ scaling; if the signatures change qualitatively, the scalar-projection shortcut fails. Alternatively, estimate the mutual information between the full marginal densities $\\rho_1(\\theta_1,\\phi_1)$ and $\\rho_2(\\theta_2,\\phi_2)$ and check whether the reported chaos signatures persist.","tokens_in":14726,"feed_emoji":"🌀","tokens_out":10565,"duration_ms":76767,"temperature":0.7,"pith_summary":"The paper argues that a purely classical quantity—mutual information between phase-space distributions of two subsystems of a kicked top—exhibits the same three chaos signatures that bipartite entanglement exhibits in the quantum version. If the argument is right, chaotic dynamics makes classical mutual information grow linearly at a rate proportional to the Lyapunov exponent, while regular dynamics makes it grow with a steadily decreasing rate; and initial conditions in chaotic regions reach a higher equilibrium mutual information than those in regular regions. The reason to care is that this would show these signatures are not intrinsically quantum: they are statistical signatures of any Liouvillian dynamics on phase space, and they expose the classical machinery behind entanglement generation.","feed_headline":"Classical mutual information mimics entanglement's chaos signature","feed_subtitle":"A purely classical measure grows at the Lyapunov rate and maps the phase space, matching quantum entropy.","key_machinery":"The carrying object is the classical kicked-top map on the unit sphere, obtained as the $j\\to\\infty$ limit of the kicked-top Hamiltonian. The paper partitions the angular momentum into a small subsystem $J_1$ (the analogue of a single spin-$1/2$) and the rest $J_2$, so the nonlinear kick term $e^{-i\\kappa(X_1+X_2)}$ couples the two sides. Nonseparability is measured by the mutual information $I_{12}$ between $X_1$ and $X_2$, estimated from sampled trajectories with a $k$-nearest-neighbor estimator, while the Lyapunov exponent is computed by a tangent-space orthogonalization algorithm. This combination—a bipartite classical map plus an information-theoretic correlation measure—converts entanglement dynamics into a classical statistical phenomenon.","core_discovery":"The central claim is that classical nonseparability, quantified by mutual information, reproduces the chaos signatures of bipartite entanglement in the kicked top. The authors bipartition the total angular momentum as $J = J_1 + J_2$, initialize a product of uniform phase-space patches around the same direction $(\\theta_0,\\phi_0)$, and evolve sampled trajectories under the classical kicked-top map. They estimate the Shannon mutual information $I_{12}$ between the rescaled components $X_1 = J_{1x}/j$ and $X_2 = J_{2x}/j$ and report three findings: $I_{12}$ grows approximately linearly at a rate of about half the positive Lyapunov exponent for chaotic initial conditions, and with a decaying rate for regular ones; the equilibration time scales as $O(\\sqrt{j})$ for regular and $O(\\ln j)$ for chaotic dynamics; and the equilibrium value $I_{12}^{\\mathrm{eq}}$ as a function of $(\\theta_0,\\phi_0)$ mirrors the classical phase-portrait structure, with higher values concentrated in chaotic regions. Each of these is a standard signature of entanglement in the quantum kicked top, so the paper offers a purely classical statistical analogue of entanglement dynamics.","pith_inferences":["A natural stress test the paper leaves open: compute $I_{12}$ between the full marginal densities $\\rho_1(\\theta_1,\\phi_1)$ and $\\rho_2(\\theta_2,\\phi_2)$ instead of the scalar components, and check whether the growth rate and equilibrium map survive.","Equal-sized partitions ($j_1 = j_2$) would reveal whether the $\\sim 0.5\\lambda$ factor and the scaling laws are general or an artifact of the strongly asymmetric split.","If the scalar-projection result is robust, it suggests a classical analogue of information scrambling in which coarse-grained mutual information growth rate plays the role of a quantum OTOC growth rate.","The authors' closing remark about charged particles in tilted magnetic wells offers a concrete experimental venue: monitor longitudinal-cyclotron energy exchange as a possible laboratory signature of classical nonseparability."],"forward_implications":["Mutual information becomes a practical classical diagnostic for chaos in phase-space distributions, complementing trajectory-based Lyapunov exponents.","The distinct equilibration-time scalings ($\\sqrt{j}$ vs $\\ln j$) give a classical statistical analogue of the speed of information spreading in regular versus chaotic dynamics.","Equilibrium mutual-information maps can be used to reconstruct classical phase-portrait structure directly from correlation data.","Because the construction uses only Liouvillian evolution, the three chaos signatures are shown to be statistical, not inherently quantum."],"supporting_citations":[{"why":"Gives the invariant inner product of phase-space distributions that motivates treating statistical states as the classical analogue of quantum states.","marker":"[1]"},{"why":"Introduces the kicked-top Hamiltonian and the classical recursion map that the paper uses for all dynamics.","marker":"[26]"},{"why":"The set of entanglement-chaos signatures (linear and logarithmic growth, equilibrium map, system-size scaling) that the paper aims to reproduce classically.","marker":"[21–25]"},{"why":"Supplies the Lyapunov-rate entanglement growth and the $O(\\sqrt{N})$ vs $O(\\ln N)$ equilibration scalings used as the quantum benchmark.","marker":"[25]"},{"why":"The $k$-nearest-neighbor estimator used to compute mutual information $I_{12}$ from sampled trajectories.","marker":"[46]"},{"why":"The tangent-space orthogonalization algorithm used to estimate positive Lyapunov exponents.","marker":"[47–49]"}],"fun_headline_variants":["Classical mutual information mimics entanglement chaos","Kicked top: classical info reproduces entanglement signatures","Phase-space mutual info traces chaos like entanglement","Mutual information in phase space acts like quantum entanglement","Classical analogue of bipartite entanglement from mutual info"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire analogy rests on the unproven assumption that the scalar mutual information between $X_1$ and $X_2$ faithfully represents the nonseparability of the full phase-space distributions $\\rho_1$ and $\\rho_2$; a different projection could give different growth rates and equilibrium maps.","fun_headline_variants_meta":{"raw":{"variants":["Classical mutual information mimics entanglement chaos","Kicked top: classical info reproduces entanglement signatures","Phase-space mutual info traces chaos like entanglement","Mutual information in phase space acts like quantum entanglement","Classical analogue of bipartite entanglement from mutual info"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000938,"raw_usage":{"total_tokens":4049,"prompt_tokens":1025,"completion_tokens":3024,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":641,"completion_tokens_details":{"reasoning_tokens":2953}},"tokens_in":641,"tokens_out":3024,"duration_ms":21552,"temperature":1.0,"reasoning_tokens":2953,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T21:14:56.417303+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute $I_{12}$ between other component pairs ($Y_1$ and $Y_2$, or $Z_1$ and $Z_2$) for the same initial conditions and compare growth rates and equilibrium maps with the $\\sim 0.5\\lambda$ scaling; if the signatures change qualitatively, the scalar-projection shortcut fails. Alternatively, estimate the mutual information between the full marginal densities $\\rho_1(\\theta_1,\\phi_1)$ and $\\rho_2(\\theta_2,\\phi_2)$ and check whether the reported chaos signatures persist.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the invariant inner product of phase-space distributions that motivates treating statistical states as the classical analogue of quantum states."},{"cited_title":"Classical and quantum chaos for a kicked top","cited_arxiv_id":null,"evidence_quote":"Introduces the kicked-top Hamiltonian and the classical recursion map that the paper uses for all dynamics."},{"cited_title":"Bridging entanglement dynamics and chaos in semiclassical systems","cited_arxiv_id":null,"evidence_quote":"Supplies the Lyapunov-rate entanglement growth and the $O(\\sqrt{N})$ vs $O(\\ln N)$ equilibration scalings used as the quantum benchmark."},{"cited_title":"https://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.mutual_info_regression.html","cited_arxiv_id":null,"evidence_quote":"The $k$-nearest-neighbor estimator used to compute mutual information $I_{12}$ from sampled trajectories."}],"review_version":1}