Oxford Quantum Circuits has launched erado, a Python library that simulates erasure noise and post-selection in Qiskit circuits. The tool enables researchers to probe error mitigation strategies before running algorithms on superconducting quantum hardware.
Post-selection is a well-known approach to error mitigation in quantum computing, but it has always faced the practical hurdle of reliably detecting and discarding corrupted results. Oxford Quantum Circuits (OQC) has now released erado, an open-source Python library that simulates erasure noise and dual-rail qubit encodings directly in Qiskit circuits. With erado, researchers can test post-selection strategies and measure their computational cost before running algorithms on actual superconducting hardware. This release comes as institutions like MIT and CERN continue to explore new methods for quantum error mitigation.
Simulating erasure and post-selection
Erado is available as a 16-qubit AerSimulator backend in OQC's Quantum Computing as a Service (QCaaS) SDK. It lets users model erasure noise-where a qubit is lost or leaves the computational state-by setting a per-gate erasure probability. The simulator supports dual-rail qubit encodings, which use two physical qubits for each logical state, and implements post-selection by discarding detected erasure events and repeating the circuit until enough valid results are collected. This method follows recent research published in Nature that emphasizes the need for realistic noise modeling in quantum algorithm development.
Users can adjust parameters such as the erasure rate (from 0.0 to 1.0 per gate), a Boolean flag for post-selection, and a false negative rate that represents the chance of missing an erasure event. The simulator also allows for idling error modeling by inserting idle gates to account for decoherence during circuit pauses. These features give researchers a controlled way to benchmark the trade-offs between error mitigation and computational overhead, a topic under active study at places like the Max Planck Society.
Numerical benchmarks and research evidence
The release is supported by a technical paper, "The limits of erasure-based postselection for quantum error mitigation" (Griffiths et al., arXiv:2606.31428), which shows that post-selection can fully mitigate the erasure channel when erasure check error rates are below 3.0%. The study finds that a postselected dual-rail system can outperform a single-rail setup at the kiloquop (1,000 quantum operations) scale, where single-rail circuits fail due to accumulated noise. This result backs the use of erasure-aware strategies for Noisy Intermediate-Scale Quantum (NISQ) algorithms, especially when full quantum error correction is not available.
Erado offers two simulation modes: CIRCUIT_SAMPLER for exact statevector sampling and TRANSPILER_PASS for memory-efficient runs on larger circuits. By tuning detector false-positive and false-negative rates, users can pinpoint where post-selection remains practical compared to standard noisy sampling. The simulator's flexible parameters allow systematic testing of error mitigation strategies under realistic noise and detection conditions, following methods used in peer-reviewed studies from Stanford and Harvard.
Broader context and competing approaches
Erasure-detection and dual-rail architectures are being explored by other quantum hardware developers as well, including D-Wave Systems (after acquiring Quantum Circuits Inc.) and Q-CTRL. By releasing erado as open source, OQC provides a reference point for benchmarking and comparison across platforms. This release follows a series of recent tools aimed at circuit optimization and error mitigation, such as the Parity Twine Optimizer for Qiskit.
Technical documentation, source code, and further analysis are available through OQC's technical blog, the arXiv preprint, GitHub, PyPI, and the QCaaS SDK documentation. Erado is intended as a research tool, not a production-ready solution. Its main value is in helping algorithm developers anticipate the impact of erasure noise and post-selection overhead before running on physical quantum processors.
Engineering limits and open questions
While erado offers a flexible simulation environment, its results depend on the accuracy of the noise and detection models. The simulator does not replace experiments on real hardware, where other error sources, calibration drift, and device variability can affect outcomes. The 3.0% erasure check error rate threshold reported in the research applies to the simulated setup and may not hold for all hardware or circuit types. The computational cost of post-selection, especially as circuits get wider and deeper, remains a practical barrier to scaling these techniques beyond small to medium-sized quantum algorithms.
Recent analysis by OQC and Trust Base, reported in 2026, suggests that raising the assumed error-correction threshold from 1% to 5% could reduce the physical qubit requirement for a 32-qubit configuration from about 404,000 to 130,000. This finding explains why dual-rail and erasure qubits are being considered as more practical fault-tolerance strategies. However, the same study notes that reaching a 5% threshold is an assumption about future hardware, not a feature of current quantum processors. For applications needing high precision (around 10-3-10-4), the number of Grover-style repetitions required would exceed current classical resource estimates, showing the significant computational overhead involved in practice.
Despite these limits, erado's open-source release helps make benchmarking of error mitigation strategies more transparent and reproducible. By exposing the trade-offs between noise, detection fidelity, and computational overhead, the tool gives researchers a clearer view of when post-selection is genuinely useful. For the quantum computing community, having access to such simulators is important for moving from theoretical error mitigation proposals to practical use on noisy, intermediate-scale devices, a challenge also noted by NASA and journals like Science.
In quantum computing, erasure noise means a qubit is lost or leaks out of the computational subspace, often due to hardware flaws or environmental effects. Post-selection is a method where measurement outcomes linked to detected errors are discarded, and the experiment is repeated until enough valid results are collected. While post-selection can suppress some errors, it increases the number of required runs and may introduce bias if detection is imperfect. Dual-rail encoding, which uses two physical qubits for one logical state, can help detect erasures but doubles the hardware needed. Balancing error suppression and resource cost is a central issue in current research on quantum error mitigation, especially for NISQ devices that lack full error correction.
