{"id":"83db9aa3-488c-46dd-9961-df9a7e3a01a1","arxiv_id":"1908.09642","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A permutation-cycle based 'integer entropy' is defined and shown to reduce to Shannon entropy in the large-N limit for the symmetric group, giving every finite group a corresponding entropy-like functional.","lead":"This paper defines a new measure of disorder, the integer entropy, based on the expected number of cycles in random permutations, and shows it approaches Shannon entropy as the number of objects grows. Scientists might care because it attaches entropy-like measures to finite symmetry groups, not just to the full set of permutations.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1's claimed limit fails when a nonvanishing fraction of the partition mass sits in bounded parts; for N=2M with partition {1,...,1,M}, lim J = I + (γ−1)/2, contradicting Eq. 4.","rationale":"The most load-bearing defect in the central claim is that the proof of Theorem 1 silently assumes every partition part contributing nonvanishing probability mass grows without bound. The counterexample with half the mass in singleton parts shows the stated limit is not equal to Shannon entropy, and the error is a nonzero constant, so it is not merely a small finite-N correction or a matter of convergence rate. This concern is exactly the first fragility identified by the reader, and it directly targets Theorem 1, the foundation of the paper. The defect is repairable by adding the missing growth hypothesis; under that hypothesis the γ terms cancel and the harmonic approximation yields Shannon entropy. Because the reader's verdict is already CONDITIONAL with this repair anticipated, my stress-test does not move the verdict. The undefined G_{n_i} for general subgroups remains a second, independent gap, but it does not change the verdict either.","tokens_in":16139,"tokens_out":12361,"duration_ms":130596,"concrete_test":"For N=2M, define partition n_1=...=n_M=1, n_{M+1}=M. Compute J=H_{2M}−1/2−(1/2)H_M exactly and Shannon I=1/2 ln(4M). Take M→∞. The difference J−I approaches (γ−1)/2≈−0.2886, not 0, which settles that Theorem 1's unqualified limit and the 'uniform convergence' statement in the abstract are false. If one instead imposes n_i→∞ for every part, the counterexample is excluded and the proof's γ-cancellation is restored.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eq. 8 gives J_{S_N}=H_N - (1/N)Σ n_i H_{n_i}. The proof of Theorem 1 approximates every H_{n_i} by ln n_i + γ and lets the γ terms cancel between H_N and the weighted sum. This cancellation is legitimate only when every n_i with nonvanishing probability mass grows without bound. The theorem and abstract state no such condition, and the claimed convergence is not uniform. For N=2M and the partition consisting of M parts of size 1 plus one part of size M, direct computation gives J_{S_N}=H_{2M}-1/2-(1/2)H_M. As M→∞, H_{2M}-(1/2)H_M = (1/2)ln N + (1/2)ln 2 + (1/2)γ + o(1), so J→(1/2)ln N + (1/2)ln 2 + (γ−1)/2. Shannon entropy of the same probabilities is I=(1/2)ln N + (1/2)ln 2. Hence lim(J−I)=(γ−1)/2≠0. Thus Eq. 4 is false as written and the abstract's 'uniformly' claim is unsupported. A repair is available: require the total mass carried by bounded parts to tend to zero, or require all n_i→∞; then the γ terms cancel and the theorem holds. This is the load-bearing gap in the central limit; the separate question of defining G_{n_i} for arbitrary subgroups remains open.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces an 'integer entropy' functional J_G for a permutation group G_N on N objects and a partition {n_i} of N, defined through expected numbers of cycles in the group and in group actions on the parts. For the symmetric group it reduces to J_{S_N}=H_N-(1/N)Σ n_i H_{n_i}. The central claim (Theorem 1, Eq. (4)) is that the N→∞ limit of J_{S_N} equals the Shannon entropy of the probability vector {n_i/N}, and the paper states this convergence is uniform. The paper then develops the Cameron–Semeraro cycle polynomial and a reciprocal 'transposition polynomial,' giving expressions for expected cycle and transposition counts in terms of polynomial roots, and illustrates the framework on alternating, cyclic, dihedral, and 'balanced' groups. An appendix also discusses broken symmetries and order parameters for the dihedral group.","tokens_in":16516,"tokens_out":8068,"duration_ms":69574,"significance":"If the central limit theorem were correct, the paper would establish a new combinatorial bridge between finite permutation groups and Shannon information, with a concrete number series for every finite group. The cycle-polynomial results (Theorems 3–6) are standard and are derived cleanly; the reciprocality between the cycle and transposition polynomials and the root-sum formulas are correct and potentially useful. The appendix computations for A_n, C_n, D_n, and balanced groups are explicit and make the framework concrete. However, as explained in the major comments, the principal limit theorem is false as stated, and the definition of the integer entropy for arbitrary subgroups is incomplete; these issues must be repaired before the paper's central claims can be accepted.","major_comments":[{"comment":"The proof of Theorem 1 replaces every harmonic number H_{n_i} by ln n_i + gamma, which is legitimate only when every n_i tends to infinity. As stated, the theorem and the abstract claim uniform convergence without this condition, and the claim is false without it. For example, take N=2M and the partition consisting of M parts of size 1 together with one part of size M; then Eq. (8) gives J_{S_N}=H_{2M}-1/2-(1/2)H_M, whose limit is (1/2)ln N + (1/2)ln 2 + (gamma-1)/2, whereas the Shannon entropy of the same probability vector is (1/2)ln N + (1/2)ln 2. The difference is the nonzero constant (gamma-1)/2. A correct statement needs an explicit growth condition, for example that the total probability mass carried by parts that remain bounded tends to zero, or that all n_i tend to infinity.","section":"§II, Theorem 1 (Eq. (4)) and proof (Eqs. (8)–(9))"},{"comment":"The formula for the integer entropy in Eq. (14a), J_G({n_i}) = <C(N)>_G - ∑ n_i <C(n_i)>_{G_i}, and its polynomial form in Eq. (14b), are missing the factor 1/N that appears in the definition Eq. (3) and in the symmetric-group expression Eq. (8). This is not a notational slip: with the factor omitted, J_{S_N} would become H_N - ∑ n_i H_{n_i}, which diverges even in simple partitions, and the appendix calculations based on Eq. (14) would be inconsistent with the main definition. The factor must be restored consistently throughout Section III and the appendix.","section":"§III, Eq. (14a) and Eq. (14b)"},{"comment":"The definition of J_G({n_i}) uses the quantities <C>_{G_{n_i}}, the expected cycle counts of the group acting on each subset of size n_i. For the symmetric group these are simply H_{n_i}, but for an arbitrary subgroup G⊆S_N there is no canonical permutation group induced on an arbitrary subset of size n_i, and the paper never defines G_{n_i}. Consequently the 'integer entropy of a finite group' is not well-defined unless one specifies, for every part of the partition, a compatible permutation representation of G on that part. This is a load-bearing gap for the claim that every finite group has a corresponding information functional; it affects not only Eq. (3) but also the examples in Appendix A.","section":"§II, Eq. (3) and §III, Eq. (14)"}],"minor_comments":[{"comment":"The name 'Cayley' is misspelled as 'Caley' in the sentence 'a result from Caley in 1849', although the reference list correctly attributes Cayley.","section":"§V"},{"comment":"Shannon entropy is denoted I in Eq. (1) but S in Eq. (9); please use a single notation throughout.","section":"§II, Eq. (1) and Eq. (9)"},{"comment":"The displayed equation (11) is garbled; the summation over group elements and the Kronecker delta are not printed clearly, making the proof harder to follow than necessary.","section":"§III, proof of Theorem 2"},{"comment":"In Table A2 and the surrounding text, the phrase 'where the d are the number of divisors of n' is awkward; it should read 'where d ranges over the divisors of n'.","section":"Appendix A.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is not ready for publication in its current form. The central limit theorem is false as stated, and the definition of integer entropy for arbitrary subgroups is incomplete. The cycle-polynomial and transposition-polynomial results are largely correct and the worked examples are instructive, so a major revision is appropriate rather than rejection. The author should also clarify the relationship to the classical fact that the expected number of cycles in S_N is H_N; this is a known result and should be cited as such in framing the paper's novelty."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: the paper has a genuinely new object — the integer entropy J_G defined through expected cycle counts over a permutation group — and the cleanly stated cycle-polynomial results are mostly correct. But the central limit theorem is not true as stated, and the advertised extension to all finite groups is not actually defined. Don't desk-reject it, but send it back for major surgery.\n\nWhat's new: the functional J_G({n_i}) = <C>_{G_N} - (1/N) Σ n_i <C>_{G_{n_i}} appears to be new, and using the Cameron-Semeraro cycle polynomial to compute it is a nice idea. Theorem 3 (logarithmic derivative gives expected cycle count) and Theorems 4-6 (root formulas for cycle and transposition averages) are correct, standard manipulations. The harmonic-number-as-expected-cycles fact is classical, but the paper's repackaging through cycle polynomials does give a useful language.\n\nThe soft spots are real. Theorem 1, the paper's centerpiece, claims a limit to Shannon entropy, with 'uniform' convergence in the abstract. The proof replaces every H_{n_i} by ln n_i + γ and lets the γ terms cancel. That cancellation is only legitimate when every part n_i grows with N. The paper states no such condition. Take N=2M with M parts equal to 1 and one part equal to M: direct calculation gives J = H_{2M} - 1/2 - (1/2)H_M, which tends to I + (γ−1)/2, not I. So Eq. 4 is false as written and 'uniform' is simply wrong. The fix is easy — require all parts to grow — but it has to be stated, and the statement of Theorem 1 has to match.\n\nThe deeper issue is that for a general subgroup G_N of S_N, the expression uses G_{n_i}, a permutation group on each subset of size n_i. A subgroup of S_N does not canonically restrict to a permutation group on an arbitrary subset; the paper never defines G_{n_i}. So the promised 'integer entropy for every finite group' is not well-defined without extra structure. This is load-bearing for the generalization, though the symmetric-group case is fine once the limit condition is repaired.\n\nOne smaller thing: Appendix B's 'balanced group' example deletes two transpositions from S_4 and claims the result is still a group. It isn't — 22 elements can't be a subgroup of S_4, and the set isn't closed. That example needs to be fixed or dropped.\n\nThe citation pattern is reasonable, and there's no data to worry about; the content is the math. The paper is worth a serious referee. The errors are specific and fixable, and the core combinatorial object may be worth keeping. I wouldn't cite it in its current form, but I'd read a careful revision.","headline":"Genuinely new combinatorial entropy whose central limit theorem is false as stated; the subgroup generalization is underdefined, but the cycle-polynomial machinery is worth a careful revision.","tokens_in":16955,"tokens_out":5603,"would_cite":false,"duration_ms":50963,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94A17","05A15","20B05","60C05"],"pacs":[],"model":"deepseek-v4-flash","headline":"The Shannon entropy of a partition is the large-N limit of an entropy defined by permutation cycle counts, and every finite group carries its own entropy.","keywords":["integer entropy","permutation cycles","Shannon entropy","finite permutation groups","harmonic numbers","cycle polynomial","transposition polynomial","symmetry breaking"],"falsifier":"A direct calculation from Eq. (8) refutes the unqualified limit: take the partition with $k$ parts of size 1 and one part of size $N-k$, with $k$ fixed, and let $N\\to\\infty$. The claimed limit, Shannon entropy, tends to $0$, while the exact expression $H_N-(kH_1+H_{N-k})/N$ tends to $\\ln N+\\gamma$, so the difference diverges; this tests the missing uniformity in Theorem 1 without any simulation.","tokens_in":15934,"feed_emoji":"🔁","tokens_out":10196,"duration_ms":104186,"temperature":0.7,"pith_summary":"This paper sets out to show that the Shannon entropy of a partition $\\{n_i\\}$ of $N$ objects is the large-$N$ limit of a purely combinatorial quantity built from permutation cycles: the integer entropy $J_{G_N}$, defined as the expected number of cycles under the group $G_N$ minus the $N$-weighted sum of the expected cycle counts inside each part. For the symmetric group $S_N$, the paper computes $J_{S_N}(\\{n_i\\})=H_N-\\frac{1}{N}\\sum_i H_{n_i}$, with $H_n$ the $n$-th harmonic number, and argues that as $N$ and every $n_i$ grow, this tends to $I(\\{n_i\\})=-\\sum_i \\frac{n_i}{N}\\ln\\frac{n_i}{N}$. The reason to care is that Shannon entropy then appears as the symmetric-group endpoint of a family of information functionals, one for every finite group, obtained by restricting the allowed permutations to a subgroup. Each such group carries its own harmonic-analog series from its cycle polynomial, so information becomes tied to symmetry rather than to a list of axioms.","feed_headline":"Permutation cycles converge to Shannon entropy","feed_subtitle":"A cycle-count-based integer entropy reproduces Shannon information in the limit and gives every finite group an entropy.","key_machinery":"The workhorse is the cycle polynomial $P_G(x)=\\sum_{\\pi\\in G}x^{c(\\pi)}$, where $c(\\pi)$ is the number of disjoint cycles in permutation $\\pi$. For the symmetric group it is the rising factorial $x^{\\overline{N}}=x(x+1)\\cdots(x+N-1)$, and the expected cycle number is the logarithmic derivative $P'_G(1)/P_G(1)$; equivalently $\\langle C\\rangle_G=\\sum_k(1-r_k)^{-1}$ with $r_k$ the roots of $P_G$. The integer entropy compares $\\langle C\\rangle_{G_N}$ with the $N$-weighted average $\\frac{1}{N}\\sum_i n_i\\langle C\\rangle_{G_{n_i}}$. The transposition polynomial $Q_G(x)=\\sum_{\\pi\\in G}x^{T(\\pi)}$, built from the Cayley relation $T(\\pi)=N-c(\\pi)$, is the reciprocal (reflected) polynomial of $P_G$, and its roots are the reciprocals of the cycle-polynomial roots.","core_discovery":"The paper's central claim is Theorem 1: in the double limit where the group becomes the full symmetric group $S_N$ and then $N$ tends to infinity, the integer entropy $J_{S_N}$ equals Shannon entropy $I(\\{n_i\\})$. The proof uses the fact that the average cycle count of a uniformly random permutation of $n$ objects is exactly $H_n$, via the signless Stirling numbers; then $J_{S_N}=H_N-\\frac{1}{N}\\sum_i H_{n_i}$, and replacing each $H_{n_i}$ by $\\ln n_i+\\gamma$ leaves $\\ln N-\\frac{1}{N}\\sum_i\\ln n_i$, which is $I(\\{n_i\\})$ because $n_i/N=p_i$. For every subgroup $G\\subseteq S_N$, the same expected-cycle construction defines an integer entropy $J_G(\\{n_i\\})$, and the logarithmic derivative of the Cameron-Semeraro cycle polynomial at $x=1$ gives the group's expected cycle number, also expressible as $\\sum_k(1-r_k)^{-1}$ over roots. The paper also introduces a reciprocal transposition polynomial whose logarithmic derivative gives the expected minimum number of transpositions needed to undo a permutation.","pith_inferences":["A renormalized version of the integer entropy that uses exact harmonic differences rather than replacing $H_{n_i}$ by $\\ln n_i+\\gamma$ would be well-defined for partitions with bounded part sizes, and comparing it with Shannon entropy across partitions would quantify how quickly the symmetric-group limit is approached.","Because $\\langle C\\rangle_G=\\sum_k(1-r_k)^{-1}$ depends only on the multiset of roots of the cycle polynomial, the integer entropy may depend on how an abstract group is embedded as a permutation group; computing it for two non-conjugate permutation representations of the same group would settle this directly.","The cyclic-group result, where expected cycle counts for prime order converge to 2 while non-prime orders show divisor-driven scatter, suggests a statistical test: for random $n$, the expected cycle count should show more variance than the harmonic series, with variance predictable from the divisor function."],"forward_implications":["Shannon entropy becomes the $S_N$ case of a wider family: restricting permutations to a subgroup $G$ yields a conditional entropy that quantifies disorder relative to that symmetry, and every finite group has such a functional.","Every finite permutation group produces a harmonic-like number series, its expected cycle count, computed from the logarithmic derivative of its cycle polynomial; the paper illustrates the alternating, cyclic, and dihedral series and shows some converge to the harmonic numbers while others, like the cyclic-group series, follow a divisor-controlled pattern.","The reciprocal transposition polynomial gives the expected minimum transposition count from the same roots, and defines balanced groups with $\\langle T\\rangle=\\langle C\\rangle=n/2$; for these, the integer entropy takes the simple form $J_B(\\mathrm{uniform})=\\frac{N}{2}\\frac{m-1}{m}$.","Because cycle structure is constant on conjugacy classes, the expected cycle numbers and the integer entropy depend only on conjugacy-class data of the group."],"supporting_citations":[{"why":"Supplies the signless Stirling numbers of the first kind and the harmonic-number interpretation of expected permutation cycles.","marker":"[18-20]"},{"why":"Presents the combinatorial identity $\\sum_k k \\left[{n\\atop k}\\right]=\\binom{n+1}{2}$, which makes the average cycle count of $S_n$ equal to $H_n$.","marker":"[19]"},{"why":"Defines the cycle polynomial of a permutation group, the object whose logarithmic derivative and roots supply expected cycle numbers for every subgroup.","marker":"[28]"},{"why":"Supplies Cayley's relation between transposition number and cycle number, $T(\\pi)=n-c(\\pi)$, underpinning the transposition polynomial and the balanced-group discussion.","marker":"[33]"}],"fun_headline_variants":["Cycle counts turn every finite group into an entropy","Shannon entropy emerges from permutation cycle limits","Integer entropy: a group-theoretic path to Shannon entropy","Permutations and symmetry: entropy's hidden group roots","Average cycle number yields a new entropy for all groups"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that every part $n_i$ grows to infinity with $N$ so that $H_{n_i}$ can be replaced by $\\ln n_i+\\gamma$, and that a subgroup $G$ of $S_N$ induces a well-defined permutation group on every subset of size $n_i$; the paper proves neither uniformly.","fun_headline_variants_meta":{"raw":{"variants":["Cycle counts turn every finite group into an entropy","Shannon entropy emerges from permutation cycle limits","Integer entropy: a group-theoretic path to Shannon entropy","Permutations and symmetry: entropy's hidden group roots","Average cycle number yields a new entropy for all groups"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000211,"raw_usage":{"total_tokens":1477,"prompt_tokens":1073,"completion_tokens":404,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":689,"completion_tokens_details":{"reasoning_tokens":330}},"tokens_in":689,"tokens_out":404,"duration_ms":4826,"temperature":1.0,"reasoning_tokens":330,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:26:30.164440+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct calculation from Eq. (8) refutes the unqualified limit: take the partition with $k$ parts of size 1 and one part of size $N-k$, with $k$ fixed, and let $N\\to\\infty$. The claimed limit, Shannon entropy, tends to $0$, while the exact expression $H_N-(kH_1+H_{N-k})/N$ tends to $\\ln N+\\gamma$, so the difference diverges; this tests the missing uniformity in Theorem 1 without any simulation.","supporting_citations":[],"review_version":1}