Native topological readout on qubit hardware: a Fibonacci-chain benchmark of measurement-compilation trade-offs

This paper benchmarks the trade-offs between native topological fusion readout and grouped-Pauli measurement strategies for Fibonacci anyon chains on NISQ hardware, revealing that the optimal method depends on the specific quantum circuit type (Floquet vs. VQE) and deriving scaling laws to guide shot-budget allocation for topological models on qubit platforms.

The Gist

The Big Question: How Do We Read a Quantum Computer's Mind?

Imagine you have a very special, complex recipe (a Hamiltonian) that describes how a magical, invisible world of particles (called Fibonacci anyons) behaves. You want to cook this recipe on a standard kitchen stove (a quantum computer made of qubits).

The problem is that your recipe uses ingredients and measurements that don't exist in a normal kitchen. In the magical world, you measure things by "fusing" particles together. But your stove only understands standard measurements, like checking if a light switch is "on" or "off" (Pauli measurements).

To get the answer, you have two choices:

1. The "Translation" Method (Grouped Pauli): You take your magical recipe, translate every single step into standard kitchen language, and then measure the switches. This is like reading a book in a foreign language by looking up every word in a dictionary as you go. It's slow and clunky, but you don't need to change the stove itself.
2. The "Native" Method (Fusion Readout): You build a special attachment for your stove that lets you measure the magical "fusion" directly. You change the state of the food right before you measure it so the stove can "see" the fusion naturally. This is like buying a special blender attachment that handles the magical ingredients perfectly.

The Paper's Goal: The author, Babatunde Moses Ayeni, wanted to know: Is it worth buying the special attachment (Native Readout), or is it better to just translate everything (Grouped Pauli)?

The answer isn't a simple "yes" or "no." It depends on how much time and energy you have.

---

The Two Scenarios Tested

The author tested these two methods on two different types of "cooking tasks":

1. The "Long March" (Digital Floquet Evolution)

• The Analogy: Imagine a long, winding hiking trail where you take thousands of small steps. The path is already mapped out; you just need to walk it.
• The Result: The Native Method (Fusion Readout) won here.
• Why? Because the trail was so long and complex, the "translation" method got bogged down in too much noise and error. The special attachment (Native Readout) was more efficient at handling the long path, giving a clearer, more accurate result with fewer mistakes.

2. The "Quick Sprint" (Optimized VQE Circuits)

• The Analogy: Imagine a very short, simple dash. You only need to take a few steps.
• The Result: The Translation Method (Grouped Pauli) won here.
• Why? Even though the special attachment (Native Readout) is better at measuring, putting it on the stove takes time and effort. On a short sprint, the time it took to attach the special tool was longer than the time saved by using it. The "translation" method was faster because it didn't require any extra setup.

---

The "Sweet Spot" (The Crossover Point)

The paper introduces a concept called the Crossover Point. Think of this like a speed limit sign on a highway.

• Below the sign (Small Budget/Short Tasks): If you have very little time or money (shots), the "Translation" method is better because it has no setup cost.
• Above the sign (Large Budget/Long Tasks): If you have a lot of time or money, the "Native" method becomes better because its superior efficiency pays off, eventually beating the translation method.

The author calculated exactly where this line is for different tasks. Sometimes the line is at the very beginning (Native is always better), and sometimes it's far down the road (Translation is better for short tasks, Native for long ones).

The "Noise" Factor

The paper also looked at what happens when the kitchen is messy (noisy hardware).

• In a perfect kitchen (Simulations): The Native method was almost always the winner because it reduced statistical errors (noise in the data).
• In a messy kitchen (Real Hardware): The Native method still reduced statistical errors, but the act of attaching the special tool introduced new errors (because the tool itself was complex and prone to glitches).
  • For the Long March, the Native method was still strong enough to win despite the messy kitchen.
  • For the Quick Sprint, the messy kitchen made the Native method's setup errors so bad that the Translation method was the clear winner.

The Bottom Line

The paper concludes that there is no single "best" way to measure quantum computers.

• If you are doing long, complex calculations (like simulating time evolution), using the Native Readout (measuring in the language of the physics) is usually worth the extra effort.
• If you are doing short, simple calculations (like finding the ground state of a small molecule), it is often better to stick to the Translation method (Pauli measurements) because the cost of setting up the Native method is too high.

The Takeaway: You shouldn't just ask, "Is this measurement method physically natural?" You must also ask, "Is the cost of setting it up worth the savings in time and error for this specific job?"

Technical Summary

Technical Summary: Native Topological Readout on Qubit Hardware

Problem Statement
Recent experimental demonstrations of non-Abelian braiding and topological order on noisy intermediate-scale quantum (NISQ) hardware have highlighted a critical design question: when does a native (topological) fusion readout genuinely offer an advantage over standard Pauli-basis reconstruction for topological anyonic Hamiltonians? While qubit processors return data in the Pauli basis, physically natural observables for topological models (such as fusion-channel projectors) often require either reconstructing them from Pauli measurements or compiling additional unitary circuits to rotate the state into a native measurement basis. The paper investigates the trade-off between the "measurement cost" (shot budget required for precision) and the "compilation cost" (circuit depth and noise introduced by basis-change unitaries) to determine when native readout survives the constraints of NISQ execution.

Methodology
The authors utilize the Fibonacci anyon chain as a minimal model to benchmark two measurement strategies:

1. Fusion Readout (FR): A native strategy where the state-preparation circuit is appended with a basis-change unitary ($U = F \cdot B$) that rotates the state into a frame where the observable is a simple local projector (direct fusion). This aligns the measurement with the Hamiltonian's structural observables.
2. Grouped Pauli (PSQWC): A baseline strategy where the observable is reconstructed from a decomposition into qubit-wise commuting (QWC) Pauli groups, without additional basis-change compilation.

The benchmark evaluates these strategies across two distinct quantum circuit families:

• Digital Floquet Evolution: Time-evolved states generated via product-formula Trotter steps.
• Optimized-State VQE: Variational Quantum Eigensolver circuits where parameters are "locked" from a noiseless simulation to isolate measurement performance from optimizer dynamics.

The performance metric is the fixed-budget Mean Squared Error (MSE) of the full energy estimator. The authors decompose the MSE into bias and variance components, specifically isolating the covariance-aware sampling variance (intrinsic to the measurement strategy) from hardware-induced noise. They derive scaling laws to identify "crossover points" ($N_c$) in the shot budget where one method becomes operationally superior to the other. Experiments were conducted on the IBM ibm_pittsburgh superconducting processor, alongside noiseless simulations.

Key Results
The analysis reveals that there is no uniform "best" method; the outcome depends heavily on the circuit regime and the balance between sampling variance and compilation overhead:

• Sampling Variance: The native Fusion Readout (FR) consistently outperforms the Grouped Pauli (PSQWC) method in terms of covariance-aware sampling variance across all tested regimes (both noiseless and hardware). This confirms that native readout reduces the intrinsic statistical uncertainty of the estimator.
• Realized MSE (Noiseless): In noiseless simulations, FR generally wins on realized MSE as well, benefiting from both lower variance and the absence of compilation errors.
• Realized MSE (Hardware): The results diverge on actual hardware:
  • Digital Floquet Circuits: FR remains favorable in the majority of cases (71/96 cells). The deep circuits inherent to Floquet evolution dilute the relative overhead of the FR basis-change layer, allowing the sampling advantage to dominate.
  • Optimized-State VQE Circuits: A sharp reversal occurs. PSQWC wins on realized MSE in 15/16 hardware cells, despite FR maintaining its sampling advantage. In these shallow variational circuits, the additional depth required for the FR basis change constitutes a significant portion of the total circuit, introducing sufficient hardware noise to outweigh the sampling benefits.
• Scaling and Crossover: The paper derives crossover points ($N_c$) where the methods switch dominance. For digital hardware, $N_c$ often lies within the sampled budget range, favoring FR at higher shot counts. For VQE hardware, the crossover points are typically very small (often below the sampled range), indicating that the compilation penalty of FR dominates the MSE even at high shot budgets in shallow circuits.

Significance and Claims
The paper claims to provide a rigorous, estimator-level framework for deciding when to measure in the native operator basis of a target physics versus falling back on Pauli-basis reconstruction. Its primary contributions are:

1. Contextualizing Native Readout: It demonstrates that while native fusion readout provides a genuine, covariance-driven sampling advantage, this advantage is not automatically preserved on NISQ hardware. The "winner" is determined by whether the sampling gain survives the non-sampling penalty of compiled basis changes.
2. Regime Dependence: The study establishes that the utility of native readout is regime-dependent. It is more likely to succeed in deep, structured circuits (like Floquet evolution) where the measurement layer is a small correction, but it may fail in shallow variational circuits (like VQE) where the measurement overhead is proportionally large.
3. Benchmarking Standard: The authors argue that evaluating measurement strategies solely on covariance or sampling variance is insufficient for NISQ hardware. The fixed-budget MSE, which accounts for both statistical efficiency and compilation-induced noise, is the necessary metric for practical decision-making.
4. Broader Relevance: While focused on Fibonacci chains, the findings extend to any topological or projector-dominated model compiled onto qubit-native platforms (e.g., superconducting or trapped-ion systems), serving as a guide for measurement design in topological engineering.

The paper concludes that native readout should not be abandoned but must be co-designed with compilation and noise considerations. It suggests that as hardware improves and logical errors are suppressed, the compilation overhead will become less of a limiting factor, potentially restoring the universal advantage of native readout.