REVIEW 3 major objections 3 minor 15 references
Sharp extremal asymptotics for Cusick's sum-of-digits bias at fixed Hamming weight
T0 review · 3 major / 3 minor · reviewed 2026-08-27 · deepseek-v4-flash
Pith's one-line read The smallest possible Cusick bias at fixed Hamming weight k is asymptotically (log₂k/k)^(3/2)/(2√π).
desk verdict Fixed-weight extremal problem with sharp constant; main theorem depends on an unverified quoted Edgeworth expansion, and the Section 9 check has a numerical bug. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing object is the critical defect $D(t)=\kappa_2(t)-\kappa_3(t)/3$, bounded below by $4$ for every $t$; here $\kappa_j(t)$ is the $j$-th cumulant of the correlation measure $\mu_t$, a coefficient that encodes the shape of that distribution. A five-cumulant Edgeworth expansion approximates $c_t-\frac12$ in powers of $M(t)^{-1/2}$. The rigidity lemmas force the cumulants of near-minimizers onto the universal ratios $(\kappa_2,\kappa_3,\kappa_4,\kappa_5)=(2,6,26,150)M+o(M)$, after which the three order-$M^{-3/2}$ correction terms cancel because their coefficients $45$, $-80$, and $35$ sum to zero. What remains is $D(t)/(8\sqrt{\pi})M^{-3/2}$, and the minimum $D=4$ produces the constant $1/(2\sqrt{\pi})$. The rigidity mechanism itself shows that near-minimizers have, up to $o(M)$ exceptions, one-blocks longer than any fixed $h$ separated by single zeros; this block geometry makes the cumulant profile universal.
What would settle it
Compute the five cumulants $\kappa_2,\ldots,\kappa_5$ and $c_t-\frac12$ exactly along the matching family $t_k=(1^{a}0\,1^{a}0\cdots 0\,1^{a})_2$ with $a\approx\log_2 k$ and $M\approx k/\log_2 k$; if the normalized ratio $(c_{t_k}-\frac12)(k/\log_2 k)^{3/2}$ does not tend to $1/(2\sqrt{\pi})$, the cancellation (4.2) or the uniformity of the Edgeworth expansion fails. Alternatively, find any $t$ with $D(t)=4$ and $M$ large for which $c_t-\frac12$ differs from $(1/(8\sqrt{\pi}))M^{-3/2}$ by more than $o(M^{-3/2})$.
Extended reading notes
Core claim
Let $s_2(n)$ be the number of ones in the binary expansion of $n$, let $c_t$ be the natural density of $n\ge0$ with $s_2(n+t)\ge s_2(n)$, and set $F(k)=\inf_{s_2(t)=k}(c_t-\frac12)$. Theorem 1.1 states that $F(k)\sim \frac{1}{2\sqrt{\pi}}\left(\frac{\log_2 k}{k}\right)^{3/2}$ as $k\to\infty$. The proof shows that any sequence approaching the infimum must have $M(t_k)\sim k/\log_2 k$ maximal one-blocks in the odd part of $t_k$, so the average one-block length is $\log_2 k$, and Theorem 1.2 adds that the critical defect $D(t)=\kappa_2(t)-\kappa_3(t)/3$ must tend to its minimum $4$ while inner zero-blocks of length at least two disappear. The constant arises because, once the cumulants take the universal profile $(\kappa_2,\kappa_3,\kappa_4,\kappa_5)=(2,6,26,150)M+o(M)$, all remaining order-$M^{-3/2}$ terms in the five-cumulant Edgeworth expansion cancel exactly, leaving $D(t)/(8\sqrt{\pi})M^{-3/2}$ as the leading contribution.
Load-bearing premise
The sharp constant rests on the five-cumulant Edgeworth expansion (2.13) being valid uniformly for all $t$ with $M(t)$ large and $D(t)$ bounded, with remainder $O(M^{-2}(\log M)^{11})$ after the auxiliary parameter is set to $R=\log M$; if that expansion has additional terms at order $M^{-3/2}$ or a larger error term, the constant $1/(2\sqrt{\pi})$ could change, and the structural first-exit lemma from the author's earlier preprint is also assumed.
Editorial extensions
If this is right
- The optimal Cusick gap at weight $k$ is polynomial-logarithmic in $k$, not exponential: it is of order $(\log_2 k/k)^{3/2}$.
- The sharp leading constant is $1/(2\sqrt{\pi})$, so the normalized ratio $F(k)(k/\log_2 k)^{3/2}$ tends to this number.
- Asymptotic extremizers have a single mesoscopic long-block scale: about $k/\log_2 k$ one-blocks of average length $\log_2 k$.
- Every integer $t$ satisfies $c_t\ge 1/2 + C(\log(2+s_2(t))/s_2(t))^{3/2}$ for an absolute constant $C>0$.
- The infinite isolated-zero model reproduces the same constant through the shadow-energy identity, giving a combinatorial certificate of the analytic cancellation.
Reading between the lines
- A natural next-order conjecture, not made in the paper, is that the correction to $F(k)$ is of order $k^{-5/2}(\log k)^{5/2}$ or similar, controlled by the sixth cumulant; nothing here rules it out.
- The rigidity theorem suggests a quantitative stability conjecture: sequences whose normalized bias lies within $\varepsilon$ of $1/(2\sqrt{\pi})$ should have $M(t_k)$ within $O(\varepsilon)$ of $k/\log_2 k$ in relative terms; this is not proved.
- The shadow-energy identity points to a purely combinatorial method that might extend to other digit bases or to the variance structure of the correlation measure; this would be a testable extension rather than a claim of the paper.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper determines the sharp asymptotic of the smallest possible Cusick bias c_t - 1/2 among integers t of fixed binary Hamming weight k. Theorem 1.1 states that F(k) ~ 1/(2 sqrt(pi)) (log_2 k / k)^{3/2}. The proof combines a first-exit lower bound, the five-cumulant Edgeworth expansion of Spiegelhofer and Wallner, and a rigidity mechanism showing that near-extremizers have M(t) ~ k/log_2 k one-blocks, isolated inner zero-blocks, and critical defect D(t) = kappa_2 - kappa_3/3 tending to 4. A matching family with isolated zeros supplies the upper bound, Theorem 1.2 gives a stability statement, and Section 9 proposes a shadow-energy interpretation of the same constant.
Significance. If correct, the result resolves the fixed-weight extremal problem with an explicit sharp constant, and it identifies a new mesoscopic block scale. The main analytical structure is coherent: the cumulant rigidity, the cancellation of the order-M^{-3/2} terms in the Edgeworth expansion, and the matching construction are mutually consistent, and the stability theorem is a natural strengthening. However, the independent shadow-energy confirmation in Section 9 is not correct as written, and the sharp constant is sensitive to the exact form and uniformity of the quoted expansion (2.13). Once those issues are addressed, the paper would be a strong contribution to the sum-of-digits literature.
major comments (3)
- [9, Eqs. (9.1) and (9.3)] Proposition 9.1 is internally inconsistent. Evaluating equation (9.1) at N=2 gives c_{t_2} - 1/2 = 1/(4*2) * binom(2,1) - (1/(12*2)) * (1 + 9) = 1/4 - 10/24 = -1/6, whereas Proposition 9.1's series (9.3) gives 1/2 * sum_{m=2}^infty 4^{-m}(m-1) = 1/18. The algebraic error occurs in the shift leading to (9.4): with A_N and B_N as defined, A_N = 1/4 (4^{-(N-1)} binom(2N-2,N-1) + B_N), so A_N - B_N/2 = 4^{-N} binom(2N-2,N-1) - B_N/4, not 1/(4N) binom(2N-2,N-1) - B_N/4. The subsequent negative-binomial tail evaluation is also mis-factored. Consequently Proposition 9.3 does not establish the claimed shadow-energy interpretation, and the agreement advertised in the introduction is unsupported. This section should be corrected or removed.
- [2.3, Eq. (2.13)] The sharp lower bound in Theorem 4.1 rests entirely on the five-cumulant expansion (2.13) and on the coefficient cancellation 45 - 80 + 35 = 0. The manuscript quotes this expansion from [13, (2.30)] without displaying the precise hypotheses, the allowed range of M, or the uniformity class in t. It only states that after choosing R = log M the conditions 'hold uniformly for all sufficiently large M'. Since Theorem 4.1 needs the expansion uniformly over all t with D(t) <= D_0 and M(t) -> infinity, the paper should either state the exact theorem from [13] and verify its conditions for this class, or provide a self-contained derivation of (2.13). A single mis-transcribed coefficient at order M^{-3/2} would change the constant, so this is a load-bearing point that needs explicit support.
- [2.2, Lemma 2.3] The few-block exclusion in Proposition 5.1 uses Lemma 2.3, whose proof depends on the exact first-exit expansion (2.3), the finite-stabilization result, the reversal identity, and the principal-subsequence representation, all from the author's unpublished preprint [2]. This is an essential step in the proof of the main theorem. Because [2] is not a published reference, the manuscript should state the needed results as lemmas with proofs, or clearly identify which statements in [2] are being assumed. As written, the paper is not fully self-contained at this load-bearing point.
minor comments (3)
- [1, final paragraph] The introduction's claim that the analytic cancellation and the shadow calculation give 'two structurally different explanations of the same extremal phenomenon' should be revised or removed until the correctness of Section 9 is established.
- [2.3, after Eq. (2.13)] The conditions on R in the quoted expansion should be stated explicitly: after taking R = log M, the requirements 8 <= R <= M^{1/6} and M^{-1/2} R <= 1/tau hold only for M sufficiently large; this should be part of the displayed statement.
- [6, proof of Proposition 6.2] The sentence 'Since M_k is asymptotic to k/A_k and A_k^2/k tends to 0, we have r_k/M_k < 1' would benefit from a one-line justification: r_k < A_k, so r_k/M_k = O(A_k^2/k).
Circularity Check
No circularity: the sharp asymptotic is derived from an external Edgeworth expansion plus new rigidity arguments; the self-cited first-exit input is prior work, not the theorem being proved.
full rationale
The paper's central derivation is not circular. The constant 1/(2√π) in Theorem 1.1 is obtained as follows: Theorem 3.5 proves the cumulant profile (κ2,κ3,κ4,κ5) ∼ (2,6,26,150)M under a bounded critical defect, using structural lower bounds and recurrence/adjacent-difference estimates quoted from Spiegelhofer and Wallner [13]; Theorem 4.1 then substitutes this profile into the quoted Edgeworth expansion (2.13) and exhibits an algebraic cancellation (45−80+35=0). No parameter is fitted to the target quantity, and no quantity is defined in terms of the claimed asymptotic. The upper-bound construction in Section 6 independently realizes the same constant by an explicit family, and Theorem 1.2 is a stability consequence, not an input. The self-citations to the author's [2] provide Lemma 2.3 and the universal lower bound (1.1); these are prior results with content different from the extremal asymptotics, and they are used only to exclude bounded-M parameters and to handle finitely many small weights. This is self-citation, not circularity, because the cited results are not equivalent to the theorem being proved and the main derivation after M→∞ stands on the external [13] machinery. Section 9 is an explicitly complementary reformulation of an exact formula from Sobolewski and Spiegelhofer [10]; it is not used in the proof of Theorem 1.1. Any concern about the validity or uniformity of the quoted expansion (2.13), or about the exact formula (9.1), is a correctness risk, not a circularity of the derivation chain.
Assumptions & free parameters
assumptions (4)
- standard math The Edgeworth expansion (2.13) from [13, (2.30)] is correct as quoted, including the error bound after taking R = log M.
- domain assumption The finite stabilization, reversal identity, and principal-subsequence representation from [2] are correct.
- standard math The structural lower bound D(t) >= 4 + K(t) + ... from [13, Corollary 2.13] and the many-block lower bound c_t > 1/2 + C M(t)^{-3/2} from [10, Remark 4] hold.
- standard math The exact formula (9.1) for the infinite isolated-zero model from [10] is correct.
Cite this review
Pith. "Pith review of Sharp extremal asymptotics for Cusick's sum-of-digits bias at fixed Hamming weight." pith.science (2026). https://pith.science/paper/PPA56FYZ
@misc{pith2026260825899,
author = {Pith},
title = {Pith review of: Sharp extremal asymptotics for Cusick's sum-of-digits bias at fixed Hamming weight},
year = {2026},
howpublished = {\url{https://pith.science/paper/PPA56FYZ}},
note = {Machine review of arXiv:2608.25899}
}
abstract
Let $s_2(n)$ be the binary sum-of-digits function and let $c_t$ be the natural density of the integers $n\ge0$ for which $s_2(n+t)\ge s_2(n)$. Earlier work of the author proved the universal exponential bound $$c_t-\frac12\ge 2^{-2s_2(t)-1},$$ thereby resolving Cusick's conjecture for every $t$. This estimate, however, does not reflect the true size of the smallest possible bias at a given large Hamming weight. In this paper, we determine this extremal scale sharply: $$\inf_{s_2(t)=k}\left(c_t-\frac12\right) \sim \frac{1}{2\sqrt\pi} \left(\frac{\log_2 k}{k}\right)^{3/2} \qquad(k\to\infty).$$ Thus the optimal fixed-weight gap is polynomial-logarithmic rather than exponential, with the explicit sharp leading constant $1/(2\sqrt\pi)$. The proof combines the five-cumulant Edgeworth expansion of Spiegelhofer and Wallner with a new extremal rigidity mechanism for near-extremal binary block patterns. We also prove a stability theorem for asymptotic extremizers and give a separate shadow-energy interpretation of the same constant.
Reference graph
Works this paper leans on
-
[2]
A first-exit proof of Cusick's sum-of-digits conjecture
K. Cheng, A first-exit proof of Cusick’s sum-of-digits conjecture, arXiv:2606.23398v2, 2026
work page Pith review arXiv 2026
-
[13]
L. Spiegelhofer and M. Wallner, The binary digits ofn+t,Ann. Sc. Norm. Super. Pisa Cl. Sci. (5)24(2023), no. 1, 1–31
work page 2023
-
[1]
J. B´ esineau, Ind´ ependance statistique d’ensembles li´ es ` a la fonction “somme des chiffres”,Acta Arith.20(1972), no. 4, 401–416
work page 1972
-
[3]
T. W. Cusick, Proof of the TuDeng conjecture, arXiv:2608.14821, 2026
work page Pith review arXiv 2026
- [4]
-
[5]
J. Emme and P. Hubert, Central limit theorem for probability measures defined by sum-of-digits function in base 2,Ann. Sc. Norm. Super. Pisa Cl. Sci. (5)19(2019), no. 2, 757–780
work page 2019
-
[6]
J. Emme and A. Prikhod’ko, On the asymptotic behavior of density of sets defined by sum-of-digits function in base 2,Integers17(2017), Paper No. A58, 28 pp
work page 2017
-
[7]
R. Liu, H. Luo and T. Xie, A complete proof for Tu–Deng conjecture, arXiv:2608.05187, 2026
work page Pith review arXiv 2026
Show all 15 references
-
[8]
J. F. Morgenbesser and L. Spiegelhofer, A reverse order property of correlation measures of the sum-of-digits function,Integers12(2012), Paper No. A47, 5 pp
2012
-
[9]
Spiegelhofer, Approaching Cusick’s conjecture on the sum-of-digits function,Integers19(2019), Paper No
L. Spiegelhofer, Approaching Cusick’s conjecture on the sum-of-digits function,Integers19(2019), Paper No. A59, 8 pp
2019
-
[10]
Sobolewski and L
B. Sobolewski and L. Spiegelhofer, Decomposing the sum-of-digits correlation measure,J. Number Theory280(2026), 702–736
2026
-
[11]
Spiegelhofer, A lower bound for Cusick’s conjecture on the digits ofn+t,Math
L. Spiegelhofer, A lower bound for Cusick’s conjecture on the digits ofn+t,Math. Proc. Cambridge Philos. Soc.172(2022), no. 1, 139–161
2022
-
[12]
Spiegelhofer and M
L. Spiegelhofer and M. Wallner, The Tu–Deng conjecture holds almost surely,Electron. J. Combin. 26(2019), no. 1, Paper No. 1.28, 28 pp
2019
-
[14]
Tar lowski, On the sum-of-digits measures and Cusick’s conjecture via stopped random walks, arXiv:2605.08624v3, 2026
D. Tar lowski, On the sum-of-digits measures and Cusick’s conjecture via stopped random walks, arXiv:2605.08624v3, 2026
2026 arXiv
-
[15]
Tu and Y
Z. Tu and Y. Deng, A conjecture about binary strings and its applications on constructing Boolean functions with optimal algebraic immunity,Des. Codes Cryptogr.60(2011), no. 1, 1–14. School of Mathematical Sciences, China West Normal University, Nanchong 637002, P. R. China Em...
2011
Reviewed August 27, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.