# US 2026/0073263 A1 · Quantum error mitigation with diffusion models

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0073263 A1
- Document ID: US20260073263A1
- Status: Published application
- Kind: A1
- Application number: 18/884,005
- Filed: 2024-09-12
- Published: 2026-03-12
- Earliest claimed: 2024-09-12
- Assignees: Dell Products L.P.
- Applicants: Dell Products L.P.
- Inventors: Rômulo Teixeira de Abreu Pinho; Miguel Paredes Quiñones; João Victor Da Fonseca Pinto; Micael Veríssimo de Araújo
- Web page: https://patentgrep.com/patent/US20260073263A1

## CPC

- G06N 10/20: Models of quantum computing, e.g. quantum circuits or universal quantum computers
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 10/70: Quantum error correction, detection or prevention, e.g. surface codes or magic state distillation
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

One example method includes performing, on a quantum circuit, a circuit cutting process to generate a set of subcircuits of the quantum circuit, for each of the subcircuits, (1) running the subcircuit on a target backend quantum hardware configuration, and measuring a noisy output resulting from the running of the subcircuit, (2) using an ML (machine learning) model to obtain an estimate of a noise-free output of the subcircuit, and (3) replacing noisy output of the subcircuits with the estimated noise-free output generated by the ML model. The method further includes running a circuit knitting process, using the estimates, to knit the subcircuits together to form the quantum circuit, and executing the quantum circuit that resulted from the circuit knitting process.

## Claims

### Claim 1

A method, comprising:
performing, on a quantum circuit, a circuit cutting process to generate a set of subcircuits of the quantum circuit;
for each of the subcircuits:
running the subcircuit on a target backend quantum hardware configuration, and measuring a noisy output resulting from the running of the subcircuit;
using an ML (machine learning) model to obtain an estimate of a noise-free output of the subcircuit; and
replacing noisy output of the subcircuits with the estimated noise-free output generated by the ML model;
running a circuit knitting process, using the estimates, to knit the subcircuits together to form the quantum circuit; and
executing, on the backend quantum hardware configuration, the quantum circuit that resulted from the circuit knitting process.

### Claim 2 (depends on 1)

The method as recited in claim 1, wherein a need to perform error mitigation after the circuit knitting process is substantially reduced relative to what the error mitigation would be in a circumstance in which the ML model is not used.

### Claim 3 (depends on 1)

The method as recited in claim 1, wherein no error mitigation is required for the subcircuits prior to performance of the circuit knitting process.

### Claim 4 (depends on 1)

The method as recited in claim 1, wherein an output of the quantum circuit, that was executed after performance of the circuit knitting process, has a lower noise than if the ML model had not been used.

### Claim 5 (depends on 1)

The method as recited in claim 1, wherein the ML model implements a diffusion modelling approach.

### Claim 6

A non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors to perform operations comprising:
performing, on a quantum circuit, a circuit cutting process to generate a set of subcircuits of the quantum circuit;
for each of the subcircuits:
running the subcircuit on a target backend quantum hardware configuration, and measuring a noisy output resulting from the running of the subcircuit;
using an ML (machine learning) model to obtain an estimate of a noise-free output of the subcircuit; and
replacing noisy output of the subcircuits with the estimated noise-free output generated by the ML model;
running a circuit knitting process, using the estimates, to knit the subcircuits together to form the quantum circuit; and
executing, on the backend quantum hardware configuration, the quantum circuit that resulted from the circuit knitting process.

### Claim 7 (depends on 6)

The non-transitory storage medium as recited in claim 6, wherein a need to perform error mitigation after the circuit knitting process is substantially reduced relative to what the error mitigation would be in a circumstance in which the ML model is not used.

### Claim 8 (depends on 6)

The non-transitory storage medium as recited in claim 6, wherein no error mitigation is required for the subcircuits prior to performance of the circuit knitting process.

### Claim 9 (depends on 6)

The non-transitory storage medium as recited in claim 6, wherein an output of the quantum circuit, that was executed after performance of the circuit knitting process, has a lower noise than if the ML model had not been used.

### Claim 10 (depends on 6)

The non-transitory storage medium as recited in claim 6, wherein the ML model implements a diffusion modelling approach.

### Claim 11

A system, comprising:
one or more hardware processors; and
a non-transitory storage medium having stored therein instructions that are executable by the one or more hardware processors to perform operations comprising:
performing, on a quantum circuit, a circuit cutting process to generate a set of subcircuits of the quantum circuit;
for each of the subcircuits:
running the subcircuit on a target backend quantum hardware configuration, and measuring a noisy output resulting from the running of the subcircuit;
using an ML (machine learning) model to obtain an estimate of a noise-free output of the subcircuit; and
replacing noisy output of the subcircuits with the estimated noise-free output generated by the ML model;
running a circuit knitting process, using the estimates, to knit the subcircuits together to form the quantum circuit; and
executing, on the backend quantum hardware configuration, the quantum circuit that resulted from the circuit knitting process.

### Claim 12 (depends on 11)

The system as recited in claim 11, wherein a need to perform error mitigation after the circuit knitting process is substantially reduced relative to what the error mitigation would be in a circumstance in which the ML model is not used.

### Claim 13 (depends on 11)

The system as recited in claim 11, wherein no error mitigation is required for the subcircuits prior to performance of the circuit knitting process.

### Claim 14 (depends on 11)

The system as recited in claim 11, wherein an output of the quantum circuit, that was executed after performance of the circuit knitting process, has a lower noise than if the ML model had not been used.

### Claim 15 (depends on 11)

The system as recited in claim 11, wherein the ML model implements a diffusion modelling approach.