Quantum Benchmark Achieves Span to Performance on Eagle, Nighthawk, Heron Processors
Summarize this article with:
Researchers are tackling a fundamental challenge in quantum computing , efficiently transferring information between distant qubits , and a new study provides a rigorous benchmark for assessing progress. Cameron V. Cogburn from the Future of Computing Institute, Rensselaer Polytechnic Institute, alongside colleagues, benchmarked inter-branch message transfer using Wigner’s-friend circuits on three leading superconducting quantum processors: Eagle, Nighthawk, and Heron (r2/r3). Their work, detailed in a new paper, implements Violaris’ unitary message-transfer primitive to analyse performance with message sizes up to a significant scale, and importantly, without relying on error mitigation. By probing device noise and routing overhead across different message types , sparse, half-weight, and dense , the team reveals crucial insights into the coherence limits of near-term quantum hardware and offers a valuable, openly-available dataset for the wider quantum community. This work establishes a rigorous method for evaluating quantum hardware’s ability to handle complex communication protocols within a multi-branch quantum system, pushing the boundaries of what’s achievable with current technology. This finding is significant as it allows for a focused assessment of inherent device limitations independent of circuit complexity. In stark contrast, the half and dense message families introduce rapidly growing routing overhead, and the variability introduced by different transpiler seeds becomes a limiting factor as the circuits approach the coherence frontier of the quantum processors. This highlights the critical interplay between circuit design, hardware connectivity, and the preservation of quantum information. These protocol stress tests provide valuable insights into the robustness of the message transfer process under varying conditions and reveal the limits of the system’s ability to maintain coherence and fidelity. The work opens avenues for understanding how complex quantum circuits degrade in performance as they scale, offering a pathway towards more efficient and reliable quantum communication protocols. All data and the scripts used for figure generation have been released, ensuring reproducibility and facilitating further research in this rapidly evolving field.
This research establishes a novel approach to benchmarking quantum processors, moving beyond traditional metrics to focus on a specific quantum communication primitive. By meticulously controlling message structure and circuit depth, the team achieved a detailed characterization of device performance across multiple architectures. The study’s emphasis on raw transfer and erasure performance, without relying on error mitigation, provides a realistic assessment of the current capabilities of superconducting quantum hardware. The findings not only advance our understanding of quantum communication but also offer a valuable benchmark suite for evaluating future quantum processors and guiding the development of more robust and scalable quantum technologies. Wigner’s-Friend Message Transfer on IBM Quantum Processors demonstrates Scientists investigated inter-branch message transfer within Wigner’s-friend circuits as a benchmark for assessing near-term superconducting processors. Researchers engineered circuits in Qiskit and executed them on IBM superconducting quantum processors via IBM Quantum cloud services, utilising the default transpiler and backend-specific coupling constraints. Experiments consistently employed 4096 shots per circuit and transpiler optimisation level 3, with error bars representing the standard deviation across compiled instances and transpiler-seed sweeps.
The team began with a pilot benchmark using a n = 1 protocol alongside two control experiments: one omitting the partial branch swap and another omitting memory uncomputation, allowing direct probing of the protocol’s efficacy and the necessity of uncomputation for erasure.
Results demonstrated the protocol successfully transferred the message with high probability, while the ‘no swap’ control failed as anticipated, and the ‘no uncompute’ control degraded memory erasure. The study then scaled the message size, analysing string transfer probability across the three message families. Conversely, half and dense messages incurred rapidly increasing routing overhead, and transpiler-seed variability became a limiting factor near the coherence frontier. Furthermore, scientists computed the bitwise mutual information I(R; Pi), averaged over active bits, to distinguish all-bits-correct string transfer from weaker correlations, finding that this diagnostic remained non-zero even when string transfer probability diminished. Table I summarises the largest message size, n, achieving a mean string-level transfer success of ≥0.1 for each family and backend, revealing ibm boston consistently achieved the largest frontier bit size, while ibm rensselaer faced limitations with half and dense messages at larger n. All data and figure-generation scripts were released to ensure reproducibility, with l μ strings recorded in the data files.
Message Transfer Benchmarks on IBM Quantum Processors demonstrate Table I summarises these findings, highlighting the performance differences across the four tested backends.
Results demonstrate that for the dense family, bitwise mutual information I(R; Pi) decayed with increasing message size n, but remained non-zero even when string transfer probability pall approached zero, indicating residual branch-paper correlations beyond strict string success. Specifically, for dense messages at n=32, I(R; Pi) remained measurable, suggesting that information was still being transferred despite complete string failure. The “twins vs cousins” benchmark, measuring branch divergence, showed that branch contrast ∆ decreased as divergence d increased, providing insights into the complexity of branch states and their impact on message transfer fidelity, the team measured a clear relationship between divergence and performance degradation, as depicted in Figure 8. All data and figure-generation scripts have been released to facilitate further research and reproducibility. 👉 More information 🗞 Inter-branch message transfer on superconducting quantum processors: a multi-architecture benchmark 🧠 ArXiv: https://arxiv.org/abs/2601.19762 Tags: