Title: Enhancing Quantum Error Correction via Adaptive Error Mitigation

Speaker: Chaithanya Naik, University of Wisconsin, Madison

Abstract: Quantum Error Correction (QEC) is essential to enable quantum computers to run quantum algorithms at practical scales. However, QEC can only detect and correct errors in two-level subspace “0" and “1". Unfortunately, qubit devices can leak to third (“2") and even higher energy levels. These leaked states are highly challenging to detect and mitigate. Moreover, if not curtailed quickly, these leakage errors spread and render QEC ineffective, resulting in logical errors. Recent experiments by Google show the severity of leakage errors during syndrome measurement and propose continuous application of leakage reduction circuits (LRCs) by reducing affected syndrome measurements due to leakage. Even these LRCs are not error-free, and unnecessarily applying LRCs increases the chance of uncorrectable errors.

On Fault-Tolerant Quantum Computer (FTQC), we will have to run QEC for millions of rounds. In this talk, we discuss HERQULES++ for hardware-efficient readout classification for leakage detection and GLADIATOR for leakage speculation to reduce leakage population growth and avoid unnecessary LRCs during QEC. GLADIATOR offers a systematic way of detecting leakage errors using a lightweight probabilistic graphical model. Moreover, our design can adapt to variable leakage rates and effectively mitigate data qubit leakage to run QEC for millions of rounds. Our evaluations show GLADIATOR lowers the false positives by 1.76× and can be deployed using a simple lookup table and reduce the leakage population growth to 2x compared to prior work.