Yue Wu

QEC Tutorial at ISCA 2026

Sun, 28 Jun 2026 00:00:00 +0000

Yue was one of the presenters of Implementing and Evaluating Quantum Error Correction Codes, a tutorial at ISCA 2026 on June 28, 2026, in Raleigh, North Carolina.

The tutorial takes a systems-oriented view of QEC, connecting logical qubits and fault-tolerant compilation with scheduling, hardware mapping, and real-time decoding. Its agenda includes a real-time decoding session and discusses decoder integration, latency, throughput, and hardware acceleration.

Invited talk: deq at QEC 2026

Mon, 08 Jun 2026 09:00:00 -0700

Yue gave an invited talk on deq at QEC 2026 in Santa Barbara, California, on June 8, 2026 (official conference schedule).

Developed during his first year at Microsoft Azure Quantum, deq is the world's first general, universal, and dynamic QEC decoding system. The talk introduced the system's support for general codes, custom logical instructions, and dynamic logical circuits with feed-forward and arbitrary branching.

Read more about the deq project, explore the open-source implementation, or follow the tutorial.

deq

Mon, 08 Jun 2026 00:00:00 +0000

Developed during my first year at Microsoft Azure Quantum, deq is the world's first general, universal, and dynamic quantum error correction decoding system. It connects declarative definitions of QEC codes and logical instructions to a runtime decoder for arbitrary dynamic logical circuits.

From LEGO proposal to deq implementation

Our earlier paper, LEGO: QEC Decoding System Architecture for Dynamic Circuits, proposed a system architecture closely matching the design now implemented in deq. It separates offline compilation into a library of decoding blocks from an online coordinator that connects logical and physical controllers to decoder backends. deq is a working, open-source system, with a language, compiler, and runtime for dynamic QEC decoding.

The architecture proposed in LEGO, closely followed by deq's implemented system. Orange components are online; blue components are offline. Figure 1 of LEGO: QEC Decoding System Architecture for Dynamic Circuits, by Yue Wu, Namitha Liyanage, and Lin Zhong (2024), used under CC BY 4.0.

Beyond a single decoding algorithm

Fast decoding of a fixed quantum memory is only part of the problem. A fault-tolerant computation changes its decoding problem as logical instructions execute, and measurement outcomes can determine which instruction runs next. deq automates the composition of these decoding problems rather than requiring an expert to hand-label an entire circuit.

General: supports general QEC codes and custom logical instruction sets through a declarative language and compiler.
Universal: supports the logical operations needed for fault-tolerant computation, including measurement-conditioned corrections.
Dynamic: handles instructions arriving at runtime, feed-forward, and arbitrary branching rather than requiring the full logical circuit in advance.
Modular: separates the system from its decoding algorithm, with built-in decoders and a stable interface for decoder plugins.
Practical: provides a shared workflow for simulation, latency benchmarking, and hardware deployment.

Hyper Blossom

Thu, 02 Oct 2025 00:00:00 +0000

Hyper Blossom (HyperBlossom in the paper) generalizes the blossom approach behind MWPM decoding from graphs to hypergraphs, making it applicable to general quantum low-density parity-check (qLDPC) codes and correlated error mechanisms.

A new graph problem for QEC

We formulate most-likely-error decoding as a minimum-weight parity factor (MWPF) problem: choose a minimum-weight set of hyperedges whose incident parities reproduce the measured syndrome. Unlike an ordinary decoding graph, a hypergraph can represent an error that triggers more than two detectors without splitting it into independent graph edges.

A decoding hypergraph, reproduced from Figure 1(b) of Minimum-Weight Parity Factor Decoder for Quantum Error Correction. Credit: Yue Wu, Binghong Li, Kathleen Chang, Shruti Puri, and Lin Zhong, 2025. Used under CC BY 4.0.

Mathematical guarantees, with explicit conditions

HyperBlossom uses a primal-dual linear-programming framework inspired by the blossom algorithm. The paper supplies mathematical proofs of optimality under stated conditions, as well as certifiable bounds on proximity to the optimum.

A feasible parity factor whose weight meets the dual lower bound is certified optimal; a remaining primal-dual gap bounds its suboptimality. General hypergraph MWPF is NP-hard, so these guarantees should not be confused with an unconditional claim of efficient exact decoding for every hypergraph.

Open-source implementation

Our implementation, Hyperion, is available in the MWPF repository, with a Rust core and Python bindings distributed as mwpf. Clustering targets almost-linear average runtime at low physical error rates.

Explore the interactive decoding demo or read the paper for the framework, proofs, and benchmarks.

Minimum-Weight Parity Factor Decoder for Quantum Error Correction

Thu, 07 Aug 2025 00:00:00 +0000

Micro Blossom

Sun, 30 Mar 2025 00:00:00 +0000

Micro Blossom is the first publicly known hardware-accelerated exact minimum-weight perfect matching (MWPM) decoder to achieve sub-microsecond decoding latency.

The FPGA prototype in our ASPLOS 2025 paper achieves 0.8 microseconds of average decoding latency for a distance-13 surface code at a physical error rate of 0.1%, running at 62 MHz. Sub-microsecond average latency is demonstrated up to distance 13 under these evaluated conditions.

This is latency, not just throughput: increasing throughput can keep pace with incoming syndrome measurements without reducing the time a dependent logical operation must wait for its correction. Micro Blossom accelerates the decoding response itself. It does not do the trick by dividing the decoding time by d rounds of syndrome extraction, which makes the data look better; rather, it faithfully report the overall latency from the last syndrome round to the time when the error-corrected logical readout becomes available. PS: The reported result is an average latency, not a worst-case guarantee for every sample.

Micro Blossom co-designs the decoding algorithm and a heterogeneous CPU/FPGA architecture, exploiting fine-grained parallelism at every vertex and edge. Figure from the Micro Blossom repository, copyright 2023 Yue Wu, used under the MIT license.

Exact decoding, hardware acceleration

The decoder retains the optimality of MWPM rather than replacing matching with an approximate heuristic. A parametrized accelerator handles massively parallel graph operations, while software handles more complex control and data structures. The implementation generates hardware for a supplied decoding graph.

Latency distributions and their effect on logical errors, from the Micro Blossom repository. The distance-9 histograms illustrate the reduction in response latency; the distance-13 headline result above is from the ASPLOS paper. Copyright 2023 Yue Wu, MIT license.

See the paper for the architecture, correctness arguments, and evaluation, and the open-source implementation for the hardware and software.

Micro Blossom: Accelerated Minimum-Weight Perfect Matching Decoding for Quantum Error Correction

Sun, 30 Mar 2025 00:00:00 +0000

LEGO: QEC Decoding System Architecture for Dynamic Circuits

Fri, 04 Oct 2024 00:00:00 +0000

LEGO is an architecture proposal, introducing decoding blocks and a modular system design for decoding dynamic logical circuits.

The architecture motivates deq, the working, open-source decoding system later developed at Microsoft Azure Quantum. LEGO lays out the high-level design; deq provides the real implementation, including its language, compiler, and runtime.

Mitigating Temporal Fragility in the XY Surface Code

Tue, 09 Jul 2024 00:00:00 +0000

Fusion Blossom: Fast MWPM Decoders for QEC

Sun, 17 Sep 2023 00:00:00 +0000

Scalable Quantum Error Correction for Surface Codes using FPGA

Sun, 17 Sep 2023 00:00:00 +0000

Fusion Blossom

Sun, 27 Aug 2023 15:53:49 -0400

Fusion Blossom combines the accuracy of exact minimum-weight perfect matching (MWPM) with almost-linear average time complexity and scalable parallel decoding. It avoids explicitly constructing the dense matching graph and enables independent regions to be decoded and then fused.

The work received the Best Paper Award at IEEE QCE 2023. Its open-source implementation provides the algorithmic foundation for later work on Micro Blossom.

Read the conference paper or explore the code.

Distributed Union-Find Decoder

Sat, 27 Aug 2022 15:53:49 -0400

This work implements a distributed Union-Find decoder on FPGA, using massively parallel processing units to achieve sub-microsecond decoding time for surface codes. The architecture is designed to scale to larger code distances.

The decoder trades the exact MWPM objective for the lower complexity of Union-Find decoding. See the IEEE QCE 2023 paper and the Helios implementation.

An interpretation of Union-Find Decoder on Weighted Graphs

Sat, 01 Jan 2022 00:00:00 +0000

Erasure conversion for fault-tolerant quantum computing in alkaline earth Rydberg atom arrays

Sat, 01 Jan 2022 00:00:00 +0000

QEC-Playground

Fri, 27 Aug 2021 15:53:49 -0400

QEC-Playground is an open-source research and educational tool for topological quantum error correction. It supports circuit simulation with Pauli and erasure noise, multiple decoders including MWPM and Union-Find, and visualization of circuits, errors, and decoding results.

Originally started as a quantum-information course project in 2020, it grew into a tool for exploring QEC and debugging decoders. The source code and documentation are available on GitHub.

Visible Light Backscatter Communication

Thu, 27 Aug 2020 15:55:35 -0400

Smart RF Table

Sat, 11 Jan 2020 21:50:31 +0800

Renovating road signs for infrastructure-to-vehicle networking: A visible light backscatter communication and networking approach

Wed, 01 Jan 2020 00:00:00 +0000

Turboboosting visible light backscatter communication

Wed, 01 Jan 2020 00:00:00 +0000

Improving visible light backscatter communication with delayed superimposition modulation

Tue, 01 Jan 2019 00:00:00 +0000

Poster: Polarization-based QAM for Visible Light Backscatter Communication

Tue, 01 Jan 2019 00:00:00 +0000

Poster: Retroreflective MIMO Communication

Tue, 01 Jan 2019 00:00:00 +0000

Long range retroreflective v2x communication with polarization-based differential reception

Mon, 01 Jan 2018 00:00:00 +0000

Software-defined visible light backscatter network

Mon, 01 Jan 2018 00:00:00 +0000

Poster: Smart RF Table Enables IoT on a Desk

Sun, 01 Jan 2017 00:00:00 +0000