💻 computer science

Hard Clique Formulas for Resolution

This paper resolves a longstanding open problem by demonstrating how to convert sparse, hard 3-CNF formulas into explicit $k$ -clique instances that are unconditionally hard to refute in Resolution, thereby establishing a conditional lower bound of $n^{\Omega(k)}$ for the problem's proof complexity.

Original authors: Albert Atserias

Published 2026-01-27

📖 3 min read☕ Coffee break read

CC BY 4.0

Original authors: Albert Atserias

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have a giant, incredibly complex puzzle made of logic rules. In the world of computer science, this is called a "3-CNF formula." Some of these puzzles are designed to be impossible to solve (unsatisfiable), and some are so tricky that even the most powerful standard solving methods (called "Resolution") take an eternity to prove they are impossible.

This paper is about taking those specific, super-hard logic puzzles and turning them into a different kind of game: the $k$ -clique problem.

The Analogy: The "Friend Group" Hunt

Think of the $k$ -clique problem like a party game. You have a room full of people (vertices), and you know who is friends with whom (edges). The goal is to find a specific group of $k$ people where everyone in that group is friends with everyone else in the group.

If $k$ is small (like 3), it's easy to find a trio of mutual friends.
If $k$ is huge (like half the room), it's incredibly hard to find that perfect circle of friends.

What the Authors Did

The researchers found a way to take a "broken" logic puzzle (one that has no solution) and translate it into a "friend group" map.

The Translation: They created a recipe to convert a hard logic puzzle into a party map. If the original logic puzzle was impossible to solve, the resulting party map will have no perfect group of $k$ friends.
The Difficulty: The magic trick is that this translation preserves the difficulty. If the original logic puzzle was exponentially hard for a computer to prove was impossible, the new "friend group" puzzle is also exponentially hard to prove impossible.
The Scale: This works for any size of the friend group ( $k$ ), as long as the group isn't too tiny or impossibly large compared to the total number of people.

Why This Matters (The "Why Should I Care?" Part)

In computer science, there is a famous guess called the Exponential Time Hypothesis (ETH). It basically says, "Some problems are just inherently slow to solve, no matter how smart your algorithm is."

The Old Way: Before this paper, we could only say, "If ETH is true, then finding these friend groups is hard." This was a conditional statement—it relied on a guess being correct.
The New Way: This paper removes the guesswork for a specific type of computer proof system (Resolution). It says, "We don't need to guess. We can unconditionally prove that these friend group puzzles are hard."

They did this by showing that the computer's proof system (Resolution) is smart enough to follow the logic of the translation they invented. Because the computer can "see" the connection, it can't cheat its way to a quick answer.

The Big Achievement

The paper solves a problem that other scientists have been stuck on for a long time (it was mentioned in literature at least twice before). They finally managed to create explicit, real-world examples of these "friend group" puzzles that are guaranteed to be incredibly difficult for computers to solve, without needing to rely on unproven theories.

In short: They built a machine that turns "impossible logic riddles" into "impossible social circle puzzles," proving once and for all that some social circles are just too complex to find, no matter how much time you spend looking.

Technical Summary: Hard Clique Formulas for Resolution

Problem Statement
The paper addresses the challenge of constructing explicit, hard instances for the Resolution proof system within the context of the parameterized $k$ -clique problem. While it is a known conditional result that, assuming the Exponential Time Hypothesis (ETH), the $k$ -clique problem cannot be solved in time $n^{o(k)}$ , the paper seeks to establish an unconditional lower bound for the length of Resolution proofs refuting the natural CNF encoding of $k$ -clique instances. Specifically, the authors aim to solve an open problem regarding the existence of explicit $k$ -clique formulas that require resolution refutations of size $n^{\Omega(k)}$ .

Methodology
The authors propose a reduction-based methodology that transforms unsatisfiable 3-CNF formulas into negative instances of the $k$ -clique problem. The core technical approach involves:

Source Selection: Starting with any unsatisfiable 3-CNF formula that is both sparse and exponentially hard to refute in Resolution.
Reduction Construction: Converting these source formulas into $k$ -clique instances. The resulting CNF encoding of these clique instances is designed to preserve the hardness properties of the original 3-CNF formulas.
Simulation Argument: The authors demonstrate that the Resolution proof system is capable of simulating the correctness proof of this specific reduction from 3-SAT to the parameterized $k$ -clique problem. This simulation capability is the linchpin that allows the transfer of hardness from the source 3-CNF formulas to the target clique formulas.

Key Contributions and Results
The paper establishes the following primary results:

Explicit Hard Instances: The authors provide a method to generate explicit negative instances of the $k$ -clique problem. For these instances, the corresponding natural CNF encoding requires Resolution refutations of size $n^{\Omega(k)}$ .
Parameter Range: This hardness result holds for any function $k = k(n)$ where $k_0 \leq k \leq n^{1/c_0}$ , with $k_0$ and $c_0$ being small constants.
Unconditional Lower Bounds: By leveraging the fact that an analogue of ETH holds for Resolution unconditionally (with explicit hard instances), the paper derives unconditional $n^{\Omega(k)}$ lower bounds for Resolution refutations of $k$ -clique. This removes the reliance on the standard ETH assumption for these specific proof complexity bounds.

Significance
The paper claims to solve an open problem that has appeared in the literature at least twice. Its significance lies in bridging the gap between conditional complexity results (based on ETH) and unconditional proof complexity results for the $k$ -clique problem. By showing that Resolution can simulate the reduction from 3-SAT to $k$ -clique, the authors re-establish the known conditional hardness result while simultaneously providing a constructive, unconditional proof of hardness for explicit instances. This confirms that the difficulty of the $k$ -clique problem is intrinsic to the Resolution system itself, independent of unproven complexity hypotheses.

The Analogy: The "Friend Group" Hunt

What the Authors Did

Why This Matters (The "Why Should I Care?" Part)

The Big Achievement

Technical Summary: Hard Clique Formulas for Resolution

More like this