IBM Quantum has introduced a new, low-level execution framework known as the directed execution model, powered by a new Executor primitive in qiskit-ibm-runtime v0.50.0. The framework transitions error mitigation logic—previously executed as a server-side black box—onto the client side, granting researchers explicit, composable control over circuit transformations, twirling, and error-correction pipelines on utility-scale hardware.
IBM Launches Directed Execution Model and Executor Primitive to Give Researchers Client-Side Control Over QEC Workflows
IBM Quantum has introduced a new, low-level execution framework known as the directed execution model, powered by a new Executor primitive in qiskit-ibm-runtime v0.50.0. The framework transitions error mitigation logic—previously…
Quantum Computing Report
Publisher
Oct 10, 2026 at 11:48 PM UTC · Updated 5時間前 · 2 分で読める

The directed execution model is designed for algorithm developers, hardware characterization experts, and quantum error correction (QEC) researchers who require custom control over large families of circuit variants. Rather than manually generating thousands of randomized circuits on the client and transferring them across the network, directed execution allows developers to define execution blueprints locally using circuit annotations and template parameter tensors, which the Executor primitive processes in the near-time environment of the IBM Quantum Compute Service.
| [ Core Components of the Directed Execution Architecture ] | ||
|---|---|---|
| Component / Tool | Software Layer & Package | Functional Role & Execution Pipeline |
| • Boxes & Annotations | • Qiskit SDK v2.0+ (BoxOp) | • Groups circuit instructions into boxes and attaches directives (e.g., Twirl, InjectNoise, ChangeBasis) for the transpiler stack. |
| • Samplomatic | • samplomatic open-source library | • Interprets annotated boxes to generate an inspectable template circuit and a structured randomization recipe (samplex). |
| • Executor Primitive | • qiskit-ibm-runtime backend engine | • Low-level execution engine that consumes QuantumProgram objects and handles shot loops, tensor parameter sweeping, and broadcasting across 5,000+ circuit randomizations. |
| • NoiseLearnerV3 | • qiskit-ibm-runtime helper program | • Client-side noise characterization tool providing sparse Pauli-Lindblad and TREX measurement noise learning for unique circuit layers. |
| • Qiskit Mitigation | • qiskit-mitigation client package | • Open-source library delivering modular routines for Probabilistic Error Cancellation (PEC), Zero-Noise Extrapolation (ZNE), and Twirled Readout Error Extinction (TREX). |
For standard users of Qiskit’s high-level Sampler and Estimator primitives, the upgrade to version 0.50.0 is non-disruptive. Both primitives now run on top of the Executor primitive as a client-side wrapper. This transition allows users to inspect and customize resilience options (such as TREX readout error mitigation) directly in local simulation modes before committing jobs to physical QPUs.
Article Intelligence
Related Coverage
Sponsored
AdNewsLayer Premium
Unlock deeper intelligence.
Ad-free reading, exclusive research, and real-time onchain insights.
Go Premium
