Exponential Lower Bounds on the Size of ResLin Proofs of Nearly Quadratic Depth

📅 2025-07-30
📈 Citations: 0
Influential: 0
📄 PDF
🤖 AI Summary
This work investigates depth–size trade-offs in the Res(⊕) proof system, aiming to overcome prior lower bounds restricted to nearly linear depth (O(N^{1+ε})). For Tseitin formulas over expander graphs, we introduce a novel lifting gadget that exhibits weak dependence on parity and satisfies secure affine masking—a property ensuring robustness under affine transformations. We further develop a lifting analysis framework applicable to arbitrary input distributions. Our main result establishes, for the first time, an exponential lower bound of exp(˜Ω(N^ε)) on proof size in Res(⊕) when the proof depth is O(N^{2−ε})—nearly quadratic in the number N of variables, for any ε > 0. This constitutes the deepest known superpolynomial lower bound for Res(⊕), substantially extending the frontier of understanding low-depth proof complexity in this system.

Technology Category

Application Category

📝 Abstract
Itsykson and Sokolov identified resolution over parities, denoted by $ ext{Res}(oplus)$, as a natural and simple fragment of $ ext{AC}^0[2]$-Frege for which no super-polynomial lower bounds on size of proofs are known. Building on a recent line of work, Efremenko and Itsykson proved lower bounds of the form $ ext{exp}(N^{Ω(1)})$, on the size of $ ext{Res}(oplus)$ proofs whose depth is upper bounded by $O(Nlog N)$, where $N$ is the number of variables of the unsatisfiable CNF formula. The hard formula they used was Tseitin on an appropriately expanding graph, lifted by a $2$-stifling gadget. They posed the natural problem of proving super-polynomial lower bounds on the size of proofs that are $Ω(N^{1+ε})$ deep, for any constant $ε> 0$. We provide a significant improvement by proving a lower bound on size of the form $ ext{exp}( ildeΩ(N^ε))$, as long as the depth of the $ ext{Res}(oplus)$ proofs are $O(N^{2-ε})$, for every $ε> 0$. Our hard formula is again Tseitin on an expander graph, albeit lifted with a different type of gadget. Our gadget needs to have small correlation with all parities. An important ingredient in our work is to show that arbitrary distributions emph{lifted} with such gadgets fool emph{safe} affine spaces, an idea which originates in the earlier work of Bhattacharya, Chattopadhyay and Dvorak.
Problem

Research questions and friction points this paper is trying to address.

Prove exponential lower bounds for ResLin proofs
Address depth constraints in ResLin proof systems
Use Tseitin formulas with expander graphs
Innovation

Methods, ideas, or system contributions that make the work stand out.

Expanding graph Tseitin formula with new gadget
Lifted gadgets with small parity correlation
Fooling safe affine spaces via lifted distributions
🔎 Similar Papers
2023-03-27Bulletin of Symbolic LogicCitations: 2