arXiv++ Combinatorics

For a prime $p\geq 3$, a somewhat restricted $3$-AP in $\mathbb{F}_p^n$ is a triplet $(x,x+a,x+2a)$, where $x\in\mathbb{F}_p^n$ and $a\in \{0,1,2\}^n$. We prove a counting lemma for somewhat restricted $3$-APs in dense sets in $\mathbb{F}_p^n$. More precisely, we prove that for all $α>0$, there exists $β>0$, such that for sufficiently large $n$, if a set $A\subseteq \mathbb{F}_p^n$ has density at least $α$, then it contains at least $β$ fraction of all somewhat restricted $3$-APs. Our proof builds on recently developed machinery from [Bhangale, Khot, Minzer, 2026]. Our main new ingredient is an arithmetic regularity lemma for patterns such as somewhat restricted 3-APs. This result is in the spirit of arithmetic regularity lemmas from the theory of Gowers uniformity norms [Green, Tao, 2010] and may be of independent interest.

We prove local and global inverse theorems for general $3$-wise correlations over pairwise-connected distributions. Let $μ$ be a distribution over $Σ\times Γ\times Φ$ such that the supports of $μ_{xy}$, $μ_{xz}$, and $μ_{yz}$ are all connected, and let $f: Σ^n \to \mathbb{C}$, $g: Γ^n \to \mathbb{C}$, $h: Φ^n \to \mathbb{C}$ be $1$-bounded functions satisfying \[ \left|\mathbb{E}_{(x,y,z) \sim μ^{\otimes n}}[f(x)g(y)h(z)]\right| \geq \varepsilon. \] In this setting, our local inverse theorem asserts that there is $δ:=\textsf{exp}(-\varepsilon^{-O_μ(1)})$ such that with probability at least $δ$, a random restriction of $f$ down to $δn$ coordinates $δ$-correlates to a product function. To get a global inverse theorem, we prove a restriction inverse theorem for general product functions, stating that if a random restriction of $f$ down to $δn$ coordinates is $δ$-correlated with a product function with probability at least $δ$, then $f$ is $2^{-\textsf{poly}(\log(1/δ))}$-correlated with a function of the form $L\cdot P$, where $L$ is a function of degree $\textsf{poly}(1/δ)$, $\|L\|_2\leq 1$, and $P$ is a product function. We show applications to property testing and to additive combinatorics. In particular, we show the following result via a density increment argument. Let $Σ$ be a finite set and $S \subseteq Σ\times Σ\times Σ$ such that: (1) $(x, x, x) \in S$ for all $x \in S$, and (2) the supports of $S_{xy}$, $S_{xz}$, and $S_{yz}$ are all connected. Then, any set $A \subseteq Σ^n$ with $|Σ|^{-n}|A| \geq Ω((\log \log \log n)^{-c})$ contains $x, y, z \in A$, not all equal, such that $(x_i,y_i,z_i) \in S$ for all $i$. This gives the first reasonable bounds for the restricted 3-AP problem over finite fields.

Let $Σ_1,\ldots,Σ_k$ be finite alphabets, and let $μ$ be a distribution over $Σ_1 \times \dots \times Σ_k$ in which the probability of each atom is at least $α$. We prove that if $μ$ does not admit Abelian embeddings, and $f_i: Σ_i \to \mathbb{C}$ are $1$-bounded functions (for $i=1,\ldots,k$) such that \[ \left|\mathbb{E}_{(x_1,\dots,x_k) \sim μ^{\otimes n}}\Big[f_1(x_1) \dots f_k(x_k)\Big]\right| \geq \varepsilon, \] then there exists $L\colon Σ_1^n\to\mathbb{C}$ of degree at most $d$ and $\|L\|_2\leq 1$ such that $|\langle f_1, L\rangle|\geq δ$, where $d$ and $δ>0$ depend only on $k, α$ and $\varepsilon$. This answers the analytic question posed by Bhangale, Khot, and Minzer (STOC 2022). We also prove several extensions of this result that are useful in subsequent applications.

We prove that any subset $A \subseteq [3]^n$ with $3^{-n}|A| \ge (\log\log\log\log n)^{-c}$ contains a combinatorial line of length $3$, i.e., $x, y, z \in A$, not all equal, with $x_i=y_i=z_i$ or $(x_i,y_i,z_i)=(0,1,2)$ for all $i = 1, 2, \dots, n$. This improves on the previous best bound of $3^{-n}|A| \ge Ω((\log^* n)^{-1/2})$ of [D.H.J. Polymath, Ann. of Math. 2012].

For a prime $p$, a restricted arithmetic progression in $\mathbb{F}_p^n$ is a triplet of vectors $x, x+a, x+2a$ in which the common difference $a$ is a non-zero element from $\{0,1,2\}^n$. What is the size of the largest $A\subseteq \mathbb{F}_p^n$ that is free of restricted arithmetic progressions? We show that the density of any such a set is at most $\frac{C}{(\log\log\log n)^c}$, where $c,C>0$ depend only on $p$, giving the first reasonable bounds for the density of such sets. Previously, the best known bound was $O(1/\log^{*} n)$, which follows from the density Hales-Jewett theorem.

We prove a stability result for general $3$-wise correlations over distributions satisfying mild connectivity properties. More concretely, we show that if $Σ,Γ$ and $Φ$ are alphabets of constant size, and $μ$ is a pairwise connected distribution over $Σ\timesΓ\timesΦ$ with no $(\mathbb{Z},+)$ embeddings in which the probability of each atom is $Ω(1)$, then the following holds. Any triplets of $1$-bounded functions $f\colon Σ^n\to\mathbb{C}$, $g\colon Γ^n\to\mathbb{C}$, $h\colon Φ^n\to\mathbb{C}$ satisfying \[ \left|\mathbb{E}_{(x,y,z)\sim μ^{\otimes n}}\big[f(x)g(y)h(z)\big]\right|\geq \varepsilon \] must arise from an Abelian group associated with the distribution $μ$. More specifically, we show that there is an Abelian group $(H,+)$ of constant size such that for any such $f,g$ and $h$, the function $f$ (and similarly $g$ and $h$) is correlated with a function of the form $\tilde{f}(x) = χ(σ(x_1),\ldots,σ(x_n)) L (x)$, where $σ\colon Σ\to H$ is some map, $χ\in \hat{H}^{\otimes n}$ is a character, and $L\colon Σ^n\to\mathbb{C}$ is a low-degree function with bounded $2$-norm. En route we prove a few additional results that may be of independent interest, such as an improved direct product theorem, as well as a result we refer to as a ``restriction inverse theorem'' about the structure of functions that, under random restrictions, with noticeable probability have significant correlation with a product function. In companion papers, we show applications of our results to the fields of Probabilistically Checkable Proofs, as well as various areas in discrete mathematics such as extremal combinatorics and additive combinatorics.

We show improved monotonicity testers for the Boolean hypercube under the $p$-biased measure, as well as over the hypergrid $[m]^n$. Our results are: 1. For any $p\in (0,1)$, for the $p$-biased hypercube we show a non-adaptive tester that makes $\tilde{O}(\sqrt{n}/\varepsilon^2)$ queries, accepts monotone functions with probability $1$ and rejects functions that are $\varepsilon$-far from monotone with probability at least $2/3$. 2. For all $m\in\mathbb{N}$, we show an $\tilde{O}(\sqrt{n}m^3/\varepsilon^2)$ query monotonicity tester over $[m]^n$. We also establish corresponding directed isoperimetric inequalities in these domains. Previously, the best known tester due to Black, Chakrabarty and Seshadhri had $Ω(n^{5/6})$ query complexity. Our results are optimal up to poly-logarithmic factors and the dependency on $m$. Our proof uses a notion of monotone embeddings of measures into the Boolean hypercube that can be used to reduce the problem of monotonicity testing over an arbitrary product domains to the Boolean cube. The embedding maps a function over a product domain of dimension $n$ into a function over a Boolean cube of a larger dimension $n'$, while preserving its distance from being monotone; an embedding is considered efficient if $n'$ is not much larger than $n$, and we show how to construct efficient embeddings in the above mentioned settings.

Given an alphabet size $m\in\mathbb{N}$ thought of as a constant, and $\vec{k} = (k_1,\ldots,k_m)$ whose entries sum of up $n$, the $\vec{k}$-multi-slice is the set of vectors $x\in [m]^n$ in which each symbol $i\in [m]$ appears precisely $k_i$ times. We show an invariance principle for low-degree functions over the multi-slice, to functions over the product space $([m]^n,μ^n)$ in which $μ(i) = k_i/n$. This answers a question raised by Filmus et al. As applications of the invariance principle, we show: 1. An analogue of the "dictatorship test implies computational hardness" paradigm for problems with perfect completeness, for a certain class of dictatorship tests. Our computational hardness is proved assuming a recent strengthening of the Unique-Games Conjecture, called the Rich $2$-to-$1$ Games Conjecture. Using this analogue, we show that assuming the Rich $2$-to-$1$ Games Conjecture, (a) there is an $r$-ary CSP $\mathcal{P}_r$ for which it is NP-hard to distinguish satisfiable instances of the CSP and instances that are at most $\frac{2r+1}{2^r} + o(1)$ satisfiable, and (b) hardness of distinguishing $3$-colorable graphs, and graphs that do not contain an independent set of size $o(1)$. 2. A reduction of the problem of studying expectations of products of functions on the multi-slice to studying expectations of products of functions on correlated, product spaces. In particular, we are able to deduce analogues of the Gaussian bounds from \cite{MosselGaussian} for the multi-slice. 3. In a companion paper, we show further applications of our invariance principle in extremal combinatorics, and more specifically to proving removal lemmas of a wide family of hypergraphs $H$ called $ζ$-forests, which is a natural extension of the well-studied case of matchings.

We survey connections of the Grothendieck inequality and its variants to combinatorial optimization and computational complexity.