Experimentally enhancing high-dimensional subchannel discrimination via optimal local filtering

The subchannel discrimination task involves distinguishing between different branches of a quantum evolution. The guessing probability of this task can be enhanced using quantum resources such as entanglement and steerability. In this study, we explore how to optimize the task's guessing probability within the one-sided device-independent quantum information processing framework, which is naturally associated with steering, and restricted to one-way local operations and classical communication. From a theoretical standpoint, we demonstrate how to separately optimize local filters and measurements, and provide an optimality condition that directly links the enhancement of guessing probability to the maximum steerability achievable through local filters. We then experimentally demonstrate the enhancement of the discrimination task on a two-ququart photonic system by implementing the optimal local filters.