Advances in Frequency-Multiplexed Readout and Subsequent Qubit-State Reset

In scalable, resource-efficient quantum processors with many superconducting qubits, readout performance often constrains overall system fidelity. Fast, high-fidelity simultaneous measurement—essential for quantum error correction—typically relies on frequency-multiplexed readout to minimize resource overhead. Yet, crosstalk and other nonidealities challenge conventional signal processing and state discrimination. Emerging machine learning approaches offer efficient, low-complexity mapping of measurement signals to qubit states, reducing error rates and enabling real-time scalability. In this talk, I will highlight recent advances in ML-based readout and reset techniques, as well as their implementation on dedicated hardware. By combining scalable algorithms with compact, ML-supported discriminators deployed on FPGAs, we can significantly contribute to overcoming the readout bottleneck, thereby enhancing both fidelity and speed.