⚛️ quantum physics

COFI-DQI: Curve-based Optimal Function Intersection via Decoded Quantum Interferometry

This paper introduces COFI, a generalization of the Decoded Quantum Interferometry (DQI) algorithm that leverages algebraic geometry codes from two-point Hermitian, Suzuki, and extended norm-trace curves to improve upon previous polynomial intersection frameworks by reducing quantum resource requirements or increasing the number of solvable constraints.

Original authors: Gretchen L. Matthews, Julia Shapiro

Published 2026-09-28

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Gretchen L. Matthews, Julia Shapiro

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

In the world of computing, there is a persistent challenge known as the maximum linear satisfiability problem. Imagine a massive spreadsheet filled with rows of instructions, where each row is a simple equation linking several variables. In a perfect world, you could find a single set of numbers for those variables that makes every single equation true. But in the messy reality of data science, engineering, and machine learning, the spreadsheet is often broken. Some rows contradict others, or the data contains errors and outliers. The goal then shifts from finding a perfect solution to finding the best possible compromise: a set of numbers that satisfies the largest number of equations possible, ignoring the few that are impossible to fix. This is a task that classical computers struggle with, especially as the number of equations grows, because the number of possible combinations to check explodes faster than any machine can handle.

To tackle this, researchers have begun looking to quantum computers, which use the strange laws of physics to explore many possibilities at once. A specific method called Decoded Quantum Interferometry has emerged as a promising tool. Think of this method as a way to turn a difficult math puzzle into a decoding problem, similar to how a radio receiver filters out static to find a clear signal. By using the mathematical structure of error-correcting codes—systems designed to fix mistakes in data transmission—this quantum approach can amplify the correct answers and suppress the wrong ones. However, for a long time, this powerful technique was limited to a narrow class of mathematical structures, much like a key that only fits one specific type of lock.

In a new study, researchers Gretchen L. Matthews and Julia Shapiro have expanded the reach of this technology. They introduced a framework they call COFI, which stands for Curve-based Optimal Function Intersection. This approach allows the quantum algorithm to work with a much wider variety of mathematical shapes, known as algebraic curves, rather than being restricted to the simple lines and circles used in previous versions. By doing so, they have shown that the quantum computer can handle more complex constraints and, in many cases, find better solutions with fewer resources. The team demonstrated that by switching to these more sophisticated curves, specifically those named Suzuki and extended norm–trace, the algorithm can satisfy a higher percentage of the equations in a system than was previously possible with the standard methods.

The core of their work involves reimagining how the quantum computer "sees" the problem. In the older approach, the computer was limited to working with simple polynomial functions, which are like basic algebraic expressions involving powers of variables. The new COFI framework allows the computer to work with rational functions, which are more flexible and can represent a wider range of behaviors. This flexibility is crucial because it lets the algorithm map the messy, real-world constraints of the satisfiability problem onto a richer mathematical landscape. The researchers proved that by using these advanced curves, the quantum algorithm can decode the "noise" in the system more effectively, leading to a higher probability of finding the optimal solution.

The study provides concrete evidence that these new curves offer tangible advantages. For instance, when comparing the new Suzuki-based approach to the previous standard, the researchers found that the new method could achieve a higher rate of satisfied equations while using fewer quantum bits, the fundamental units of information in a quantum computer. In some scenarios, the improvement was significant enough to allow the system to handle a larger number of constraints without requiring a massive increase in computing power. The team also explored two-point Hermitian codes, another variation of these curves, and found that they too could outperform the older one-point versions, particularly in situations where the system was not yet fully saturated with constraints.

One of the most practical findings concerns the efficiency of the hardware. The researchers calculated that using these new curves reduces the number of quantum bits needed to represent each piece of data. In the context of quantum computing, where building and maintaining qubits is one of the biggest engineering hurdles, this reduction is vital. It means that for the same amount of physical hardware, a quantum computer using the COFI framework could solve larger and more complex problems than one using the older, more limited methods. The study does not claim to have solved the satisfiability problem for all cases, but it establishes a clear path forward, proving that the quantum advantage is not limited to a single type of mathematical structure.

The work also includes a direct comparison with a well-known classical algorithm called Prange's algorithm. In the tests performed, the quantum approach consistently outperformed the classical method, finding solutions that satisfied a greater fraction of the equations. This gap in performance was not just a theoretical possibility; the researchers provided specific numerical examples where the quantum method showed a clear edge, even with relatively small field sizes. This suggests that the quantum advantage is robust and can be realized in practical settings, not just in idealized mathematical models.

By broadening the class of curves that can be used, the researchers have opened the door for future improvements. The study suggests that the potential for optimization is not fixed but depends on the choice of the underlying mathematical family. As the field of quantum computing matures, the ability to select the most efficient curve for a given problem could become a standard tool for engineers and scientists. The findings indicate that the future of quantum optimization lies not in a single magic bullet, but in a diverse toolkit of mathematical structures, each tailored to extract the maximum performance from the quantum hardware.

Ultimately, this paper marks a significant step in making quantum optimization more practical and powerful. It moves the field beyond the initial, limited demonstrations and shows that by leveraging the deep geometry of algebraic curves, we can build quantum algorithms that are both more efficient and more effective. The results provide a clear roadmap for how to construct these systems, offering a way to handle the complex, noisy data that defines modern science and industry. As quantum computers continue to evolve, the ability to navigate these mathematical landscapes will likely become a cornerstone of their utility, turning what was once a theoretical curiosity into a reliable engine for solving the world's most difficult optimization problems.

Technical Summary: COFI-DQI: Curve-Based Optimal Function Intersection via Decoded Quantum Interferometry

Problem Statement
The paper addresses the Maximum Linear Satisfiability (max-LINSAT) problem over finite fields $\mathbb{F}_q$ . Given a system of $n$ linear equations in $k$ variables, the goal is to find an assignment that satisfies the maximum possible number of equations, even in the presence of noise or inconsistencies. This problem is a finite-field specialization of the maximum feasible subsystem problem, known to be computationally hard (NP-hard) and difficult to approximate beyond random assignment baselines for general instances.

The authors focus on leveraging the Decoded Quantum Interferometry (DQI) framework, a quantum algorithm introduced by Jordan et al. (2025), to achieve provable speedups on structured instances. DQI reduces optimization problems to decoding problems for classical error-correcting codes. Specifically, it maps the objective function of max-LINSAT to a syndrome-decoding problem for the dual of a linear code. The performance of DQI depends on the decoding radius of the dual code; a larger decoding radius allows for a higher degree of polynomial amplification, leading to a higher expected fraction of satisfied constraints.

Methodology
The paper introduces COFI (Curve-based Optimal Function Intersection), a generalization of the Optimal Polynomial Intersection (OPI) and Hermitian Optimal Polynomial Intersection (HOPI) frameworks. While OPI and HOPI restrict optimization to polynomial functions over the projective line or the Hermitian curve, respectively, COFI extends this to the Riemann-Roch space $L(G)$ of divisors on arbitrary smooth projective curves over $\mathbb{F}_q$ .

The methodology proceeds as follows:

Framework Generalization: The authors define COFI as finding a function $f \in L(G)$ that maximizes the number of points $P_i$ where $f(P_i)$ falls within a specified constraint set $F_i \subseteq \mathbb{F}_q$ . This allows for optimization over rational functions, not just polynomials, by utilizing algebraic geometry (AG) codes.
Dual Code Analysis: The DQI algorithm requires efficient unique decoding of the dual code $C^\perp$ . The authors analyze the duals of AG codes derived from specific curve families. They rely on the property that the dual of an AG code is also an AG code from the same curve, allowing the application of efficient decoding algorithms (e.g., Berlekamp-Massey-Sakata, Feng-Rao, or interpolation-based methods) to the dual.
Satisfaction Fraction Derivation: Using the semi-circle law derived in previous DQI literature, the authors derive a lower bound for the expected satisfaction fraction $s/n$ . This bound is a function of the decoding radius $\ell$ of the dual code, the field size $q$ , and the number of constraints $n$ . The key metric is the ratio $\ell/n$ , which determines the algorithm's ability to bias the quantum state toward optimal solutions.
Specific Curve Families: The paper applies this framework to three specific families of curves with many rational points:
- Suzuki Curves: Defined over $\mathbb{F}_q$ with $q=2^{2t+1}$ .
- Extended Norm-Trace Curves: Defined over $\mathbb{F}_{q^t}$ , which generalize the Hermitian curve.
- Two-Point Hermitian Codes: Utilizing divisors supported at two points rather than one.

Key Contributions

Introduction of COFI: The authors establish a unified framework for applying DQI to optimization problems over Riemann-Roch spaces of arbitrary curves, broadening the scope beyond polynomial representations.
Theoretical Bounds: They derive explicit formulas for the expected DQI satisfaction fraction for AG codes from general curves, linking the bound directly to the genus $g$ and the degree of the divisor $G$ .
Performance Improvements:
- Suzuki Codes: The paper demonstrates that one-point Suzuki codes can achieve improved satisfaction fractions while requiring fewer qubits per field element compared to Hermitian codes of comparable length.
- Extended Norm-Trace Codes: Under specific assumptions on the divisor degree, these codes attain larger satisfaction fractions than corresponding one-point Hermitian codes.
- Two-Point Hermitian Codes: The authors show that a class of two-point Hermitian codes yields a strictly larger expected satisfaction fraction than one-point Hermitian codes of the same dimension, provided the comparison lies in the nonsaturated regime.
Comparison with Classical Baselines: The paper compares DQI performance using Suzuki codes against Prange's algorithm (a classical randomized decoding algorithm), showing that DQI achieves a higher expected satisfaction fraction even at small field sizes.

Results

Qubit Efficiency: By utilizing curves with many rational points (e.g., Suzuki and extended norm-trace), the authors show that the number of constraints $n$ can be increased relative to the field size $q$ . This reduces the number of qubits required to represent a single field element compared to Reed-Solomon codes (which have $n \le q$ ) and, in some cases, Hermitian codes.
Satisfaction Fractions: Numerical examples and theoretical comparisons (e.g., Example 4.5 and Table 4) illustrate that for matched block lengths and constraint densities, Suzuki codes and two-point Hermitian codes provide strictly higher lower bounds on the expected satisfaction fraction than one-point Hermitian codes.
Decoding Radius: The analysis confirms that the dual codes of these specific curve families possess decoding radii that, when normalized by the code length, exceed those of the standard one-point Hermitian construction in specific parameter regimes.

Significance
The paper claims that COFI extends the applicability of the DQI framework beyond the previously studied Reed-Solomon and one-point Hermitian settings. By leveraging the rich algebraic structure of algebraic geometry codes from curves of positive genus, the authors demonstrate that it is possible to improve the theoretical guarantees of quantum optimization algorithms. Specifically, the work shows that selecting appropriate curve families can reduce quantum resource requirements (qubits per field element) or increase the number of constraints that can be effectively processed, thereby offering a pathway to more efficient quantum solutions for combinatorial optimization problems. The authors note that these improvements are contingent on the existence of efficient unique decoding algorithms for the dual codes, a condition met by the specific families analyzed.

More like this