Finite version of the $q$-analogue of de Finetti's theorem

Adyan Dordzhiev

REVIEW 4 major objections 3 minor 7 references

Finite version of the $q$-analogue of de Finetti's theorem

T0 review · 4 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Finite q-exchangeable laws sit within c_k q^n of mixtures

desk verdict The intended finite q-analogue of de Finetti may be a real question, but this draft's proof collapses on a wrong inversion-count exponent, and the paper's own definition of q-exchangeability is inconsistent with its extremal measures. read the letter →

arxiv 2506.01176 v1 pith:Q6QEKDBL submitted 2025-06-01 math.PR math.CO

Adyan Dordzhiev This is my paper

classification math.PRmath.CO MSC 60G0905A30

keywords q-exchangeabilitydeFinettitheoremfiniteexchangeablesequencestotalvariationdistanceq-binomialcoefficientssharprateextrememeasuresq-Bernoulli

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proves a finite version of the $q$-analogue of de Finetti's theorem. For any $q$-exchangeable probability measure on $\{0,1\}^n$, there is a mixture of $q$-Bernoulli measures that agrees on the first $k$ coordinates up to variational error at most $c_k q^n$, with $c_k$ depending only on $k$. The paper also shows the exponential rate $q^n$ is optimal. If correct, the result gives quantitative, finite-$n$ control parallel to the classical finite de Finetti bound $4k/n$, and it recovers the infinite $q$-analogue as a limit. The proof proceeds through the convex structure: every $q$-exchangeable measure is a convex combination of $n+1$ extreme measures, and each extreme measure is compared explicitly with a single $q$-Bernoulli measure.

What carries the argument

The key mechanism is the convex decomposition of the compact space of $q$-exchangeable measures into extreme points $e_{n,i}^q$, together with the explicit formula (2.3.3) for the projection of an extreme measure onto the first $k$ coordinates. That formula writes $(e_{n,n_1})_k(s_{k,k_1})$ as a ratio of $q$-binomial coefficients times a power of $q$. The proof of Lemma 3.1.3 reduces the variational distance to estimating this ratio against the $q$-Pochhammer term $(q^{n_1}; q^{-1})_{k_1}$, splitting into the cases $k_1 = k$ and $k_1 < k$. The $q$-binomial identities carry the error into a factor $q^n$ multiplied by a constant depending only on $k$.

What would settle it

Compute both sides of equation (2.3.3) directly for a small quadruple such as $n=3$, $k=1$, $n_1=2$, $k_1=0$: the left side is a sum of two weights of the extreme measure over the possible one-bit continuations, while the right side is the claimed $q$-binomial ratio. The theorem's quantitative conclusion depends on the equality holding for all admissible quadruples, so a single mismatch is enough to settle the matter.

Watch

Extended reading notes

Core claim

The central claim is Theorem 3.1.1: for $q \in (0,1)$, every $q$-exchangeable probability measure $P$ on $\{0,1\}^n$ admits a probability measure $\mu$ on $\Delta_q = \{1,q,q^2,\ldots\} \cup \{0\}$ such that for every $k \le n$, the total variation distance between the projection of $P$ to the first $k$ coordinates and the projection of the mixture $P_{\mu,k}$ is at most $c_k q^n$. Section 3.4 constructs an extreme example showing that no rate faster than $q^n$ can hold uniformly, so the order is optimal. The paper therefore asserts that finite $q$-exchangeable laws are, up to exponentially small projection error, exactly mixtures of $q$-Bernoulli laws.

Load-bearing premise

The load-bearing premise is the explicit formula (2.3.3), which identifies the projection of an extreme $q$-exchangeable measure as $q^{(n_1-k_1)(k-k_1)}\binom{n-k}{n_1-k_1}_q / \binom{n}{n_1}_q$; if this displayed identity is incorrect, Lemma 3.1.3 and Theorem 3.1.1 lose their stated support.

Editorial extensions

If this is right

For fixed $k$, the approximation error on the first $k$ coordinates decays exponentially in $n$, so choosing $n$ of order $\log(1/\varepsilon)$ guarantees accuracy $\varepsilon$.
The optimality result rules out any uniform upper bound of the form $c_k q^{n+\delta}$ with $\delta > 0$.
The theorem re-derives the infinite $q$-analogue of de Finetti's theorem by compactness: finite approximations plus a subsequence limit give the integral representation (2.1.7).
The constants $c_k$ are produced explicitly by the proof, up to crude inequalities, so the bound is not merely an existence statement.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

For practical simulation, the theorem suggests that a $q$-exchangeable process can be approximated by first sampling an atom $x$ from $\Delta_q$ according to $\mu$ and then generating a $q$-Bernoulli sequence; the error on $k$-marginals after $n$ steps is at most $c_k q^n$.
The same convex-decomposition route could produce finite-$n$ quantitative analogues for other deformed exchangeabilities, provided the extreme measures admit projection formulas comparable to (2.3.3).
The sharpness example indicates that the uniform rate is governed by the extreme measures themselves, so no averaging over mixtures can beat $q^n$.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The one thing you should know: the central theorem is not proven. Proposition 2.3.3, which is load-bearing for Lemma 3.1.3 and Theorem 3.1.1, has the wrong exponent. The question—a finite q-analogue of de Finetti with rate q^n—is legitimate and worth asking, but this draft does not answer it.

What is new and what works: the finite formulation with claimed sharp rate q^n is not in Gnedin-Olshanski or Diaconis-Freedman, so the statement is new. The convexity approach, proving the extreme case and then mixing, is a sensible strategy. The q-binomial identity in Proposition 2.1.3 is correct, and the paper is clearly organized.

The soft spot is fatal. In Proposition 2.3.3, the marginal of the extreme measure e_{n,n1} on the first k coordinates, evaluated at the sequence with k1 ones then zeros, has inversion-count exponent k1(n−n1), not (n1−k1)(k−k1). The proof's inversion count for s_{k,k1} ∪ s_{n−k,n1−k1} drops the within-prefix inversions and the crossings between prefix ones and tail zeros; the correct total exponent simplifies to k1(n−n1). With the correct exponent, Lemma 3.1.3 is false. For example, take q=1/2, n=10, n1=2, k=2. The first-coordinate marginal of e_{10,2} is exponentially close to a point mass at 0, while ν_{q^2} gives P(X1=1)=1−q^2=3/4, so the variation distance is a constant around 1.5, not O(q^10). The sharpness argument in Section 3.4 inherits the same error.

There is also an internal inconsistency: Definition 1.0.2 as printed contradicts the extreme measures in Proposition 2.3.1. For q=1/2, n=2, e(10)=1/3 and e(01)=2/3, but (1.0.3) would force P(01)=q P(10)=1/6. The text's accompanying interpretation suggests a slipped sign in (1.0.3), but as written the definition and the paper's own objects disagree.

Who this is for: people working on q-exchangeability will care about the intended statement, but the current draft is not a reliable source. It deserves a serious referee because the result is important and the error, though fatal, is the kind of algebraic slip that might be repaired by choosing a different approximating measure. As is, I would not cite it. If you are the editor, send it to review with the expectation of rejection or major revision; the author needs to prove the extreme-case bound with a correctly chosen parameter, not just fix the exponent.

Referee Report

4 major / 3 minor

Summary. The paper claims a finite version of the q-analogue of de Finetti's theorem: for any q-exchangeable probability measure P on {0,1}^n, there exists a probability measure µ on the discrete set Δ_q such that the total variation distance between the k-dimensional marginals of P and of the mixture P_{µ,n} is at most c_k q^n, with c_k depending only on k. The proof proceeds by decomposing P into extreme q-exchangeable measures e_{n,n1}, pairing each extreme measure with the point mass δ_{q^{n1}}, and estimating the resulting marginal distance via an explicit formula for the pushforward of e_{n,n1} (Proposition 2.3.2). Sharper lower bounds are claimed in Section 3.4, and a corollary rederives the infinite Gnedin–Olshanski theorem.

Significance. If correct, the result would be a natural q-analogue of the Diaconis–Freedman finite de Finetti theorem, with the interesting feature that the approximation error decays exponentially in n rather than polynomially. The convex-geometric strategy is well motivated and the paper is clearly written in its overall structure. However, the central technical result rests on an incorrect algebraic identity, and a direct counterexample shows that the key lemma is false. Given those defects, the paper in its current form does not establish its main theorem.

major comments (4)

[Section 2.3, Proposition 2.3.2, Eq. (2.3.3)] The exponent (n1-k1)(k-k1) in formula (2.3.3) is incorrect. For the concatenated word s_{k,k1} ∪ ω' with ω' containing n1-k1 ones and n-k-(n1-k1) zeros, the number of inversions is inv(s_{k,k1}) + inv(ω') + k1(n-n1-k+k1). Summing over ω' gives the prefactor q^{k1(n-n1)} times [n-k choose n1-k1]_q, not q^{(n1-k1)(k-k1)}. For a direct check, take n=3, n1=2, k=1, k1=1: the left side of (2.3.3) equals (q+q^2)/(1+q+q^2), while the right side equals (1+q)/(1+q+q^2). This error propagates into Lemma 3.1.3 and Sections 3.2 and 3.4.
[Section 1.0.2 and Section 2.1, Eqs. (1.0.3), (2.1.1), Prop. 2.3.1(2)] The definitions of q-exchangeability are inconsistent with the extreme measures listed in Proposition 2.3.1(2). For q=1/2 and n=2, the measure e_1 defined by e_1(10)=q^1/[2]_q=1/3 and e_1(01)=1/[2]_q=2/3 does not satisfy (1.0.3), because that relation would require e_1(01)=q e_1(10)=1/6. The same inconsistency appears in (2.1.1), where the sign of the exponent is reversed relative to the formula e_q^k(ω)=q^{inv(ω)}/[n choose k]_q. Consequently, the convex decomposition used in the proof of Theorem 3.1.1 is not a decomposition into q-exchangeable measures under the paper's own stated definition.
[Section 3.1, Lemma 3.1.3] Lemma 3.1.3 is false. Let q=1/2, n=2m, n1=m, and k=1. Using the correct marginal from the formula above, (e_{2m,m})_1(1)=q^m [m]_q/[2m]_q, which tends to 0 as m→∞. On the other hand, the measure ν_q^{q^m} in (2.1.5) gives (ν_q^{q^m})_1(1)=1-q^{m-1}, which tends to 1. The total variation distance between the first-coordinate marginals is therefore about 1, whereas the lemma would require it to be at most c_1 q^{2m} with c_1 independent of m. This directly contradicts (3.1.4) and invalidates the proof of Theorem 3.1.1.
[Section 3.2, Eq. (3.2.5) and subsequent estimates] The proof of Lemma 3.1.3 relies on the erroneous factor q^{(n1-k1)(k-k1)} in (3.2.1). With the correct factor q^{k1(n-n1)}, the term k1=0 carries no suppression at all, and the difference between the q-binomial ratio and (q^{n1};q^{-1})_{k1} is not uniformly O(q^n) in n1. The uniform constant c_k claimed in (3.1.2) and (3.1.4) cannot be obtained from the estimates (3.2.7)–(3.2.11). Thus the central rate bound is unsupported.

minor comments (3)

[Section 2.3, notation] The notation for extreme measures is not consistent: Proposition 2.3.1(2) uses e_q^k, while Section 2.3 and later sections write e_{n,n1}^q. Please unify the notation and define the dependence on n and n1 explicitly.
[Section 2.1, Eq. (2.1.4)] The summation in (2.1.4) is written as "sum over C_{n,k}" without the explicit index; the intended expression is sum_{ω∈C_{n,k}} q^{inv(ω)} = [n choose k]_q. Please correct the display.
[Throughout] The document contains numerous typographical artifacts (e.g., "/greaterorequalslant" in Section 3.2 and 3.4, "ﬁn ite" in the introduction). These should be cleaned before submission.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the finite q-de Finetti theorem is derived from finite extreme measures and q-binomial estimates, with no fitted parameters or author self-citation chain.

full rationale

The paper's central derivation chain is not circular in the pith sense. Theorem 3.1.1 is proved by decomposing an arbitrary q-exchangeable finite measure into a convex combination of the finite extreme measures e^q_i from Proposition 2.3.1(2), defining the mixing measure µ by setting µ(q^i)=α_i, and then bounding each summand via Lemma 3.1.3. Lemma 3.1.3 is proved directly from Proposition 2.3.2 and explicit q-binomial estimates; it does not invoke the infinite Gnedin–Olshanski theorem as an input. The infinite theorem is cited for context and for the infinite extreme-point identification in Proposition 2.3.1(1), but the proof of the finite theorem does not depend on that identification. Section 3.3 uses the finite theorem to derive the infinite surjectivity result, which is the reverse direction from circularity. No parameter is fitted to a subset of data and then renamed a prediction, no uniqueness theorem is imported from the authors' own prior work, and no ansatz is smuggled in via citation. The apparent defects flagged in the reader's take — the inversion-count exponent in equation (2.3.3) and the consistency of the proposed extreme measures with Definition 1.0.3 — are mathematical or algebraic errors that affect the correctness of the proof, but they do not make the derivation circular. The proof is self-contained relative to external benchmarks, so the appropriate circularity score is 0.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The paper introduces no free parameters or invented entities. The proof is analytic and relies on standard convexity and q-binomial facts; the failure is an algebraic error in Proposition 2.3.3, not an unstated assumption.

assumptions (3)

domain assumption The set of q-exchangeable probability measures on {0,1}^n is convex and compact.
Used in Section 2.3 to apply Choquet's theorem and to decompose any such measure into a convex combination of extreme points.
standard math Choquet's theorem: a compact convex set is the closed convex hull of its extreme points.
Invoked in Section 2.3 to justify the extreme-point decomposition used in the proof of Theorem 3.1.1.
domain assumption The infinite q-analogue de Finetti representation (Gnedin-Olshanski) holds and Delta_q is compact.
Used in Section 3.3 for the corollary from finite to infinite; the main theorem statement only defines the target mixtures, but the corollary re-derives the known result.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Finite version of the $q$-analogue of de Finetti's theorem." pith.science (2026). https://pith.science/paper/Q6QEKDBL

@misc{pith2026250601176,
  author       = {Pith},
  title        = {Pith review of: Finite version of the $q$-analogue of de Finetti's theorem},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Q6QEKDBL}},
  note         = {Machine review of arXiv:2506.01176}
}

abstract

Let $q \in (0,1)$. We formulate an asymptotic version of the $q$-analogue of de Finetti's theorem. Using the convex structure of the space of $q$-exchangeable probability measures, we show that the optimal rate of convergence is of order $q^n$.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

7 extracted references · 7 canonical work pages

[2]

Diaconis and D

P. Diaconis and D. Freedman. Finite Exchangeable Sequences . The Annals of Probability , 8(4):745 -- 764, 1980

work page 1980
[1]

Representations of the Infinite Symmetric Group

Alexei Borodin and Grigori Olshanski. Representations of the Infinite Symmetric Group . Cambridge Studies in Advanced Mathematics. Cambridge University Press, 2016

work page 2016
[3]

W. Feller. An Introduction to Probability Theory and Its Application Vol II . John Wiley and Sons, 1971

work page 1971
[4]

A q-analogue of de Finetti's theorem

Alexander Gnedin and Grigori Olshanski. A q-analogue of de Finetti's theorem . The Electronic Journal of Combinatorics , Volume 16, Issue 1 (2009)

work page 2009
[5]

q-exchangeability via quasi-invariance

Alexander Gnedin and Grigori Olshanski. q-exchangeability via quasi-invariance. The Annals of Probability , 38(6), November 2010

work page 2010
[6]

Edwin Hewitt and Leonard J. Savage. Symmetric Measures on Cartesian Products . Transactions of the American Mathematical Society , 80(2):470--501, 1955

work page 1955
[7]

An elementary proof of de Finetti's Theorem , 2018

Werner Kirsch. An elementary proof of de Finetti's Theorem , 2018

work page 2018

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [2]

Diaconis and D

P. Diaconis and D. Freedman. Finite Exchangeable Sequences . The Annals of Probability , 8(4):745 -- 764, 1980

work page 1980

[2] [1]

Representations of the Infinite Symmetric Group

Alexei Borodin and Grigori Olshanski. Representations of the Infinite Symmetric Group . Cambridge Studies in Advanced Mathematics. Cambridge University Press, 2016

work page 2016

[3] [3]

W. Feller. An Introduction to Probability Theory and Its Application Vol II . John Wiley and Sons, 1971

work page 1971

[4] [4]

A q-analogue of de Finetti's theorem

Alexander Gnedin and Grigori Olshanski. A q-analogue of de Finetti's theorem . The Electronic Journal of Combinatorics , Volume 16, Issue 1 (2009)

work page 2009

[5] [5]

q-exchangeability via quasi-invariance

Alexander Gnedin and Grigori Olshanski. q-exchangeability via quasi-invariance. The Annals of Probability , 38(6), November 2010

work page 2010

[6] [6]

Edwin Hewitt and Leonard J. Savage. Symmetric Measures on Cartesian Products . Transactions of the American Mathematical Society , 80(2):470--501, 1955

work page 1955

[7] [7]

An elementary proof of de Finetti's Theorem , 2018

Werner Kirsch. An elementary proof of de Finetti's Theorem , 2018

work page 2018