Lower Bounds on Query Complexity for Testing Bounded-Degree CSPs

In this paper, we consider lower bounds on the query complexity for testing CSPs in the bounded-degree model. First, for any ``symmetric’’ predicate $P:{0,1}^{k} \to {0,1}$ except \equ where $k\geq 3$, we show that every (randomized) algorithm that distinguishes satisfiable instances of CSP(P) from instances $(|P^{-1}(0)|/2^k-\epsilon)$-far from satisfiability requires $\Omega(n^{1/2+\delta})$ queries where $n$ is the number of variables and $\delta>0$ is a constant that depends on $P$ and $\epsilon$. This breaks a natural lower bound $\Omega(n^{1/2})$, which is obtained by the birthday paradox. We also show that every one-sided error tester requires $\Omega(n)$ queries for such $P$. These results are hereditary in the sense that the same results hold for any predicate $Q$ such that $P^{-1}(1) \subseteq Q^{-1}(1)$. For EQU, we give a one-sided error tester whose query complexity is $\tilde{O}(n^{1/2})$. Also, for 2-XOR (or, equivalently E2LIN2), we show an $\Omega(n^{1/2+\delta})$ lower bound for distinguishing instances between $\epsilon$-close to and $(1/2-\epsilon)$-far from satisfiability. Next, for the general k-CSP over the binary domain, we show that every algorithm that distinguishes satisfiable instances from instances $(1-2k/2^k-\epsilon)$-far from satisfiability requires $\Omega(n)$ queries. The matching NP-hardness is not known, even assuming the Unique Games Conjecture or the $d$-to-$1$ Conjecture. As a corollary, for Maximum Independent Set on graphs with $n$ vertices and a degree bound $d$, we show that every approximation algorithm within a factor $d/\poly\log d$ and an additive error of $\epsilon n$ requires $\Omega(n)$ queries. Previously, only super-constant lower bounds were known.

💡 Research Summary

This paper investigates the fundamental limits of query complexity for property testing of constraint satisfaction problems (CSPs) in the bounded‑degree model. The authors focus on two main axes: (1) the nature of the predicate defining the CSP, and (2) the distance parameter that separates satisfiable instances from those that are far from satisfiable.

For any symmetric predicate (P:{0,1}^k\to{0,1}) with (k\ge 3) (excluding the equality predicate EQU), they prove that any randomized tester that must distinguish perfectly satisfiable instances from instances that are (\bigl(|P^{-1}(0)|/2^k-\varepsilon\bigr))-far from satisfiable requires at least (\Omega!\bigl(n^{1/2+\delta}\bigr)) queries, where (\delta>0) depends only on (P) and (\varepsilon). This improves upon the classic “birthday‑paradox” lower bound of (\Omega(\sqrt{n})). The proof proceeds by analyzing the probability that two randomly sampled variables share a constraint, establishing a non‑negligible collision probability, and then showing that with fewer than (n^{1/2+\delta}) queries an adversarial distribution can hide the global unsatisfiability while appearing locally consistent.

In the one‑sided error setting (the tester may only err on rejecting a satisfiable instance), the authors obtain a much stronger (\Omega(n)) lower bound for the same class of predicates. The argument uses an information‑theoretic encoding of the hidden assignment and shows that any tester that queries fewer than a linear number of variables cannot recover enough bits to guarantee a one‑sided guarantee.

For the equality predicate EQU, they complement the lower bounds with a one‑sided tester that runs in (\tilde{O}(\sqrt{n})) queries, demonstrating that the hardness is indeed predicate‑specific. They also treat the 2‑XOR (or E2LIN2) predicate, showing an (\Omega\bigl(n^{1/2+\delta}\bigr)) lower bound for distinguishing (\varepsilon)-close from ((1/2-\varepsilon))-far instances, again highlighting that even linear equations over (\mathbb{F}_2) inherit the same query barrier.

Moving beyond specific predicates, the paper establishes a universal linear lower bound for general binary‑domain (k)-CSPs. Any algorithm that must separate satisfiable instances from those that are ((1-2k/2^k-\varepsilon))-far requires (\Omega(n)) queries. This result is notable because the corresponding NP‑hardness of approximation (even under the Unique Games or (d)-to‑1 conjectures) is not known, indicating a gap between algorithmic hardness and property‑testing hardness.

As a concrete application, the authors translate the linear lower bound to the Maximum Independent Set problem on graphs with maximum degree (d). They prove that any algorithm achieving an approximation factor of (d/\operatorname{polylog} d) while also guaranteeing an additive error of (\varepsilon n) must make (\Omega(n)) queries. Prior work only yielded super‑constant lower bounds for this setting, so this result substantially strengthens our understanding of query complexity for graph approximation in the bounded‑degree model.

Overall, the paper delivers a comprehensive suite of lower bounds that break the previously believed (\Omega(\sqrt{n})) barrier for many natural CSPs, establishes linear‑query hardness for a broad class of problems, and clarifies the delicate interplay between predicate symmetry, one‑sided error, and distance parameters in property testing. These insights set a new baseline for future algorithmic work aiming to approach or circumvent these fundamental query limits.