The Signal
IBM researchers model hybrid quantum jobs as workflows
An IBM Research workshop paper models hybrid quantum-classical applications as directed acyclic graphs and formulates their scheduling as a mixed-integer linear programming problem. It evaluates the approach with Sample-Based Quantum Diagonalization and QAOA, reporting that optimized scheduling can improve resource use and reduce end-to-end execution time.

Scheduling across a whole workflow
A new IBM Research workshop paper models hybrid quantum-classical applications as directed acyclic graphs, then formulates their scheduling as a mixed-integer linear programming problem. The paper, dated November 15, 2026, is listed for SC 2026. Its five authors are Vadim Elisseev, Evgeny Shindin, Odellia Boni, Orit Davidovich and Vasileios Kalantzis; IBM identifies them as affiliated with the company at publication.
The approach treats an application as coordinated quantum and classical tasks running across different resources. Rather than optimizing an individual quantum or classical component alone, the framework addresses scheduling at the workflow level.
Dependencies, costs and uncertainty
The scheduling formulation accounts for task dependencies and assignment to heterogeneous resources, as well as communication costs, makespan and tardiness. The paper also develops a robust counterpart based on budgeted uncertainty to account for the stochastic nature of quantum processing unit execution times.
These elements are part of the stated optimization problem: the graph represents the application’s task structure, while the scheduling formulation considers where tasks run and the costs and timing measures named in the paper.
Evaluation with two algorithms
The researchers evaluate the method using Sample-Based Quantum Diagonalization and the Quantum Approximate Optimization Algorithm. IBM Research says the evaluation illustrates that optimized scheduling can improve resource utilization and reduce end-to-end execution time in hybrid quantum-classical environments.
It presents the work as a workshop paper, not as a released scheduling product or a claim that a particular execution-time reduction has already been achieved in production.