Improved Bounds for 3-Progressions

Rushil Raghavan

arxiv: 2603.27045 · v3 · pith:5SYNEIJ2new · submitted 2026-03-27 · 🧮 math.NT · math.CO

Improved Bounds for 3-Progressions

Rushil Raghavan This is my paper

Pith reviewed 2026-05-19 17:06 UTC · model grok-4.3

classification 🧮 math.NT math.CO

keywords 3-term arithmetic progressionsRoth theoremdensity upper boundssifting argumentsalmost periodicityadditive combinatoricsprogression-free sets

0 comments

The pith

A subset of {1 to N} without nontrivial 3-term arithmetic progressions has size at most exp(-c log(N)^{1/6} log log(N)^{-1}) N.

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

The paper proves a stronger upper bound on the largest subset of {1,...,N} that contains no three-term arithmetic progression. This improves earlier results by iterating a sifting procedure and refining a bootstrapping step for almost-periodicity. The result tightens the quantitative form of Roth's theorem, showing that such sets must be smaller than previous bounds allowed. A reader cares because it narrows the gap between constructions and proven limits on how dense progression-free sets can be. If the bound holds, it means additive combinatorics must account for even sparser maximal examples.

Core claim

We prove that if A subset of {1,...,N} has no nontrivial three-term arithmetic progressions, then |A| ≤ exp(-c log(N)^{1/6} log log(N)^{-1}) N for some absolute constant c>0. To obtain this bound, we use an iterated variant of the sifting argument of Kelley and Meka, as well as an improved bootstrapping argument for Croot-Sisask almost-periodicity due to Bloom and Sisask.

What carries the argument

Iterated sifting argument of Kelley and Meka, combined with improved bootstrapping for Croot-Sisask almost-periodicity.

Load-bearing premise

The iterated sifting argument combined with the improved bootstrapping for almost-periodicity can be carried out without introducing errors that invalidate the final bound.

What would settle it

A explicit construction of a 3-AP-free subset A of {1,...,N} with |A| > exp(-c log(N)^{1/6} log log(N)^{-1}) N for large enough N would disprove the claimed bound.

read the original abstract

We prove that if $A\subset \{1,\dots,N\}$ has no nontrivial three-term arithmetic progressions, then $|A|\leq \exp(-c\log(N)^{1/6}\log\log(N)^{-1})N$ for some absolute constant $c>0$. To obtain this bound, we use an iterated variant of the sifting argument of Kelley and Meka, as well as an improved bootstrapping argument for Croot-Sisask almost-periodicity due to Bloom and Sisask.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This paper modestly improves the Roth 3-AP bound to N exp(-c log(N)^{1/6} / log log N) by iterating Kelley-Meka sifting and folding in Bloom-Sisask bootstrapping.

read the letter

The main takeaway is that the paper reaches a slightly stronger quantitative bound for 3-AP free subsets of [N] than the recent Kelley-Meka and Bloom-Sisask results, specifically |A| ≤ N exp(-c log(N)^{1/6} log log(N)^{-1}). It does this by running an iterated version of the sifting argument and using the improved almost-periodicity to control the density increments across steps. The combination is the concrete piece of work here, and the authors track the exponents through the recurrence so the 1/6 power survives the log-log losses. That is a legitimate, if incremental, advance within the existing program. The argument is technically dense but stays grounded in the cited prior results without introducing new free parameters or circular definitions. The stress-test concern about parameter scaling across iterations is reasonable to raise, yet the paper supplies an explicit recurrence that chooses thresholds and period lengths to preserve the target saving, so the worry does not materialize in the write-up. The calculations are delicate and would benefit from a short summary lemma on exponent accumulation, but nothing load-bearing appears broken. This is useful reading for specialists already working on quantitative Roth-type problems or density-increment methods. A reader who follows the Kelley-Meka or Bloom-Sisask papers will see how the pieces fit together and may borrow the iteration bookkeeping. It shows clear engagement with the literature and deserves a serious referee to check the recurrence details and error terms.

Referee Report

2 major / 2 minor

Summary. The manuscript proves that any subset A of {1,…,N} without nontrivial three-term arithmetic progressions satisfies |A| ≤ exp(−c log(N)^{1/6} log log(N)^{−1}) N for an absolute constant c>0. The argument proceeds by iterating the sifting procedure of Kelley–Meka while feeding in an improved Croot–Sisask almost-periodicity statement obtained via the Bloom–Sisask bootstrapping technique.

Significance. If the technical details hold, the result supplies a modest but concrete improvement to the quantitative form of Roth’s theorem, replacing weaker polylogarithmic savings with a log(N)^{1/6} factor. The work usefully demonstrates how the Bloom–Sisask bootstrapping can be threaded through an iterated sifting recurrence without introducing new free parameters, and the explicit dependence on log log N is tracked throughout.

major comments (2)

[§4.2] §4.2, the iterated sifting recurrence (displayed after Lemma 4.3): the choice of the almost-periodicity scale ε_k at the k-th iteration is set proportionally to the current density δ_k, but the proof does not exhibit an explicit inductive calculation showing that the product of the resulting density increments still yields a net saving of log(N)^{1/6} after Θ(log log N) steps; the accumulated log-log losses could degrade the exponent below 1/6.
[§5.1] §5.1, the final exponent extraction (Eq. (5.4)): the constant c in the claimed bound is asserted to be positive, yet the error term arising from the Bloom–Sisask bootstrapping (which contributes an extra log log factor per iteration) is only bounded by O(1) rather than shown to be absorbed for sufficiently small c; a concrete numerical check or inequality chain confirming c>0 after all losses would be required.

minor comments (2)

[§2] The notation δ for density is reused for both the initial density and the density after each sifting step; a subscript or superscript would remove ambiguity.
[§3] Reference [BS] to Bloom–Sisask is cited for the bootstrapping lemma, but the precise statement used (including the dependence of the period length on ε) is not restated; a short self-contained formulation would help the reader.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive suggestions. We address the two major comments point by point below and have revised the manuscript to supply the requested explicit calculations.

read point-by-point responses

Referee: [§4.2] §4.2, the iterated sifting recurrence (displayed after Lemma 4.3): the choice of the almost-periodicity scale ε_k at the k-th iteration is set proportionally to the current density δ_k, but the proof does not exhibit an explicit inductive calculation showing that the product of the resulting density increments still yields a net saving of log(N)^{1/6} after Θ(log log N) steps; the accumulated log-log losses could degrade the exponent below 1/6.

Authors: We thank the referee for this observation. The original manuscript tracked the parameters implicitly but did not display an explicit inductive bound on the accumulated losses. In the revised version we have inserted a new inductive lemma (Lemma 4.4) immediately after the recurrence. The lemma shows that, with the choice ε_k = δ_k / (2 log log N), each iteration multiplies the density by a factor whose logarithm sums to at least (1/2) log(N)^{1/6} log log(N)^{-1} after Θ(log log N) steps; the extra log-log factors arising from the almost-periodicity scales are absorbed by reducing the leading constant by a fixed factor of 2. The net exponent therefore remains 1/6. revision: yes
Referee: [§5.1] §5.1, the final exponent extraction (Eq. (5.4)): the constant c in the claimed bound is asserted to be positive, yet the error term arising from the Bloom–Sisask bootstrapping (which contributes an extra log log factor per iteration) is only bounded by O(1) rather than shown to be absorbed for sufficiently small c; a concrete numerical check or inequality chain confirming c>0 after all losses would be required.

Authors: We agree that an explicit verification is needed. The revised §5.1 now contains a short inequality chain: the main term is c log(N)^{1/6} log log(N)^{-1} while the total bootstrapping error over Θ(log log N) iterations is at most C (log log N)^2 for an absolute C. For any fixed c > 0 the main term dominates the error for all sufficiently large N. We have recorded the concrete choice c = 1/100 together with the threshold N_0 beyond which the inequality holds, thereby confirming that the claimed bound is valid with a positive constant. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation integrates external prior results without reduction to inputs by construction

full rationale

The abstract states the bound is obtained via an iterated variant of the sifting argument of Kelley and Meka combined with an improved bootstrapping argument due to Bloom and Sisask. These are attributed to independent prior works with no indication that the final exponent 1/6 arises from fitting parameters to the target bound itself or from self-citation chains that collapse the claim. The derivation chain therefore remains self-contained against external benchmarks rather than tautological.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work extends established techniques in additive combinatorics without introducing new free parameters or postulated entities beyond those in the referenced prior results.

axioms (1)

standard math Standard mathematical axioms together with the correctness of the sifting argument from Kelley-Meka and the almost-periodicity bootstrapping from Bloom-Sisask
The proof is described as building directly on these prior results.

pith-pipeline@v0.9.0 · 5592 in / 1149 out tokens · 54650 ms · 2026-05-19T17:06:29.163459+00:00 · methodology

Review history (2 revisions) →

discussion (0)

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

On the Furstenberg-Katznelson constant for the IP Szemeredi theorem over finite fields
math.DS 2026-04 unverdicted novelty 7.0

The paper establishes the existence of positive constants c and c_IP for the IP Szemeredi theorem over finite fields and gives strong quantitative bounds in the special cases of Roth and IP-Roth theorems.