| Abstract: | The quantum noise inherent in contemporary quantum processors possesses a rich and complex structure that depends heavily on the specific computation. In this work, we exploit local time-reversal operations and insights into noise accumulation to reliably amplify noise in a controllable manner. Counterintuitively, this amplification can be leveraged to mitigate errors. More generally, quantum error mitigation (QEM) techniques are the engine that facilitates the extraction of meaningful results from today's noisy intermediate-scale quantum devices. While QEM methods based on noise characterization are vulnerable to non-stationary noise drifts during an experiment, we demonstrate that our approach is fully resilient to such variations. This provides unprecedented robustness, allowing for the mitigation of significantly higher noise levels. Furthermore, our framework is compatible with mid-circuit measurements and the resulting partial collapse of the quantum state. Our findings are supported by experimental data from both superconducting and trapped-ion quantum computers. Finally, we outline future research directions. |