⚛️ quantum physics

Practical Fidelity Estimation for Noisy Quantum Devices Using Direct and Survival-Probability-Based Protocols

This paper introduces two practical, software-oriented protocols for estimating quantum fidelity on noisy intermediate-scale devices without full tomography: an importance-sampled direct fidelity estimation method enhanced with attenuation-aware corrections, and a robust survival-probability-based approach utilizing $L_2$ -optimized expansions to mitigate SPAM errors and outperform standard truncation in high-noise regimes.

Original authors: Ahmed Aman, Kaustav Chatterjee, Younes Javanmard, Tanmoy Pandit

Published 2026-10-16

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ahmed Aman, Kaustav Chatterjee, Younes Javanmard, Tanmoy Pandit

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

Building a computer that uses the strange rules of quantum physics to solve problems requires a way to check if the machine is actually working. In the world of quantum computing, information is stored in delicate units called qubits. Unlike the bits in a standard laptop, which are either zero or one, qubits can exist in a mix of both states at once. However, these states are incredibly fragile. The slightest disturbance from heat, vibration, or electromagnetic waves can cause the information to degrade or disappear entirely. To build a useful quantum computer, scientists must be able to measure how well a device preserves information, a quality known as fidelity. Traditionally, checking this required a process called tomography, which is like trying to map the entire surface of a complex, spinning globe by taking millions of individual photographs from every possible angle. As the number of qubits grows, this method becomes impossibly slow and expensive, requiring more measurements than there are atoms in the universe for even moderately sized machines. This bottleneck has forced researchers to look for smarter, faster ways to gauge performance without needing to see every single detail of the quantum state.

A team of researchers has developed two new software-based methods to solve this problem, offering practical tools for testing noisy quantum devices without the heavy burden of full reconstruction. The first approach, known as direct fidelity estimation, works by taking a targeted guess rather than a complete survey. Instead of measuring every possible property of the quantum state, the method uses a statistical trick to focus only on the most important parts of the state that are known in advance. Imagine trying to estimate the average height of a crowd by measuring only the tallest people, weighted by how likely they are to appear; this is the essence of the technique. The researchers found that this method is highly effective when the target state is known, but it is sensitive to errors in the measurement process itself. To fix this, they introduced a correction that accounts for the fact that quantum measurements are often slightly "dimmed" by noise, much like a camera lens that loses a bit of light. By mathematically compensating for this dimming, the team showed that the estimates become much more accurate, even when the measurement equipment is imperfect.

The second method takes a completely different path, focusing on how the system behaves over time rather than its static structure. This approach, called survival-probability estimation, tests the machine by sending a known initial quantum state forward through a noisy process and then trying to reverse that process to see if the state returns to its original form. The researchers measure how often the state survives this round trip. If the machine were perfect, the state would always return; if it is noisy, the state will drift away. The key innovation here is how the data is analyzed. A simple mathematical guess based on small errors works well when the noise is low, but it fails miserably when the noise is strong. The team discovered that by using a more sophisticated mathematical optimization to fit the data, they could accurately reconstruct the fidelity even in very noisy conditions. This optimized approach proved to be far superior to standard methods in high-noise environments, providing a reliable diagnostic tool where other techniques would give up.

Both methods were tested extensively using computer simulations that mimicked real-world quantum devices. The researchers simulated various types of noise, including random flips of the qubit states and gradual loss of information, across systems with multiple qubits. They found that the direct estimation method, when corrected for measurement dimming, could accurately predict the quality of the state with high confidence. Similarly, the survival-probability method, using the optimized mathematical fit, successfully tracked the degradation of the system even when the noise was severe. The simulations covered a range of scenarios, from simple random states to complex, structured states, confirming that both tools are robust and adaptable. The results suggest that these protocols can serve as complementary tools for engineers: one is best when the target state is well understood and structured, while the other excels when the goal is to monitor the cumulative effect of noise over time, provided the initial state is known and the experimental cycle matches specific theoretical conditions.

The significance of this work lies in its practicality for the current era of quantum technology, often called the noisy intermediate-scale quantum era. In this phase, devices are large enough to be interesting but still too noisy to run complex algorithms without errors. The ability to quickly and accurately measure performance is essential for debugging these machines and improving their design. The researchers emphasize that these are not just theoretical ideas but software-oriented protocols that can be implemented on existing hardware. While the current results are based on simulations, the framework is designed to be tested on real superconducting quantum processors. The team plans to deploy these methods on actual hardware to see how they hold up against real-world imperfections like calibration drift and correlated errors. By providing a flexible toolbox for fidelity estimation, this research helps bridge the gap between characterizing a device and optimizing its control, paving the way for more reliable and scalable quantum computers in the future.

Technical Summary: Practical Fidelity Estimation for Noisy Quantum Devices Using Direct and Survival-Probability-Based Protocols

Problem Statement
Reliable characterization of quantum states and processes is fundamental to the development of scalable quantum computing. However, standard techniques like full quantum state tomography become experimentally and computationally impractical as system size increases due to unfavorable scaling costs. There is a critical need for task-specific estimators that can determine operational figures of merit, such as fidelity, without reconstructing the full density matrix. While existing methods like Direct Fidelity Estimation (DFE) and periodic-driving diagnostics offer partial solutions, a unified framework comparing their specific advantages, limitations, and noise sensitivities across different benchmarking scenarios has been lacking.

Methodology
The paper presents and analyzes two distinct, software-oriented fidelity estimation protocols designed for Noisy Intermediate-Scale Quantum (NISQ) devices:

Direct Fidelity Estimation (DFE):
- Principle: This protocol estimates the overlap between a known target state $\rho$ and an experimentally prepared state $\sigma$ using importance-sampled Pauli measurements. By expanding states in the normalized Pauli basis, the fidelity is expressed as a sum of products of characteristic functions.
- Sampling: Pauli operators are selected according to an importance-weighted distribution $p(k) = X_\rho(k)^2$ , where $X_\rho(k)$ are the target state's Pauli coefficients.
- Noise Handling: The authors analyze the impact of readout noise, demonstrating that measurement errors attenuate Pauli expectation values. They propose an "attenuation-aware" correction where measured values are divided by a calibrated attenuation factor $\beta_k = (1-2\alpha)^{w(k)}$ (where $w(k)$ is the weight of the Pauli string) to recover unbiased estimates.
Survival-Probability-Based Protocol:
- Principle: This approach avoids explicit Pauli reconstruction. Instead, it characterizes the cumulative effect of a noisy evolution (represented by a superoperator $K$ ) by measuring return probabilities (survival probabilities) after repeated forward–inverse cycles.
- Objective: The goal is to reconstruct an "adjoint-cycle fidelity proxy" $F_Q = \langle\langle \rho_0 | \sqrt{Q} | \rho_0 \rangle\rangle$ , where $Q = K^\dagger K$ . This proxy approaches the physical fidelity if the implemented inverse operation realizes the adjoint evolution $K^\dagger$ and the error map is positive semidefinite.
- Reconstruction: The fidelity proxy is reconstructed from moments $R_m = \langle\langle \rho_0 | C^m | \rho_0 \rangle\rangle$ (where $C$ is the forward-inverse cycle) using a polynomial expansion $F_Q \approx \sum c_m R_m$ .
- Optimization: The authors argue that standard Taylor expansion coefficients are only optimal in weak-noise regimes. They propose an $L_2$ -norm optimization to determine expansion coefficients over a physically relevant spectral interval $[g, 1]$ , which provides a more robust approximation in high-noise regimes.

Key Contributions

Unified Framework: The paper combines DFE and survival-probability methods into a comparative benchmarking framework, clarifying their complementary nature. DFE relies on the structural knowledge of a target state's Pauli representation, while the survival-probability method relies on dynamical information from evolution cycles.
Attenuation Correction for DFE: The authors provide a clear software-level formulation of DFE and demonstrate that applying attenuation-factor corrections significantly improves robustness against measurement noise, maintaining accuracy even with error rates up to ~5%.
$L_2$ -Optimized Polynomial Reconstruction: For the survival-probability approach, the paper formulates a reconstruction method using $L_2$ -optimized coefficients. Numerical evidence shows this approach substantially outperforms standard Taylor truncation in moderate-to-strong noise regimes.
Noise Sensitivity Analysis: The work analyzes the theoretical basis and noise sensitivity of both protocols, specifically addressing depolarizing, dephasing, and amplitude-damping channels.

Results
Numerical simulations were performed on representative multiqubit states (Haar-random, GHZ, and W states) under various noise models:

DFE Performance: In simulations of four-qubit Haar-random states under global depolarizing noise, the mean estimated fidelity closely matched the true fidelity when attenuation corrections were applied. The distribution of estimates remained stable even with readout noise.
Survival-Probability Performance: For two-qubit states under depolarizing noise, the $L_2$ -optimized reconstruction of the adjoint-cycle fidelity proxy showed marked improvement over Taylor-based approximations as noise strength increased.
Limitations: The survival-probability simulations utilized exact moments and did not include finite-shot fluctuations or explicit State-Preparation-and-Measurement (SPAM) errors. Consequently, while the method shows theoretical promise for SPAM robustness, its practical resilience to these specific experimental imperfections remains to be established experimentally.

Significance and Claims
The paper claims to offer a flexible toolbox for fidelity estimation suitable for near-term quantum devices, provided specific conditions are met:

DFE is presented as an efficient, statistically rigorous route for states where the Pauli structure is known and efficiently computable, particularly for small-to-intermediate size demonstrations.
Survival-Probability Method is presented as a complementary, dynamics-based approach useful where repeated return experiments are accessible, provided the implemented cycle matches the theoretical adjoint cycle ( $Q = K^\dagger K$ ) and spectral assumptions hold.

The authors emphasize that these methods are not competing replacements but complementary tools addressing different experimental constraints. They conclude that while numerical simulations validate the theoretical robustness of the proposed corrections and optimizations, the next critical step is the deployment of these protocols on real superconducting quantum hardware (specifically citing VTT's infrastructure) to validate their performance against established techniques like randomized benchmarking and to assess resilience to calibration drifts, leakage, and correlated noise.

More like this