[ASPLOS 2026] ReloQate: Beating Quantum Noise Drift via Real-Time "Reloqation"

TL;DR

Quantum hardware is not a static environment—noise levels drift constantly due to material defects and cosmic rays. ReloQate introduces a dynamic "operating system" layer for Surface Code QEC that monitors the Detector Fire Rate (DFR) to predict logical failures. When a qubit tile starts to "sour," the system hot-swaps the logical information to a clean, calibrated tile (Remapping), allowing the original tile to be fixed without stopping the computation.

The Motivation: The "Static Noise" Fallacy

Most Quantum Error Correction (QEC) theories treat physical error rates as fixed constants. In reality, superconducting and trapped-ion qubits are subject to Temporal Drift. A qubit that worked perfectly at 9:00 AM might be useless by 9:05 AM due to a Two-Level System (TLS) fluctuation.

Prior works like CaliScalpel or SurfDeformer attempted to solve this using Patch Deformation—growing the qubit's footprint to route around "bad" physical qubits. However, these methods are often:

1. Statically Scheduled: They assume we know when drift happens (we don't).
2. Resource Hungry: They require massive "buffer zones" of physical qubits around every logical qubit.

Methodology: From Parity to Prediction

The core insight of ReloQate is that the Detector Fire Rate (DFR)—the frequency at which stabilizer measurements flip—is a high-fidelity proxy for the Logical Error Rate (LER).

1. The DFR-to-LER Mapping

The authors demonstrate a power-law relationship: $p_L\propto (extDFR{)}^{(d+1)/2}$ By maintaining a sliding window (buffer) of recent DFRs, the system can calculate a "Confidence Interval" for the current LER. This allows the controller to be Zealous (responding early to avoid any breach) or Sluggish (saving resources but risking errors).

Figure 1: The ReloQate pipeline—monitoring DFR buffers, predicting LER breaches, and triggering a remap to a fresh tile.

2. The Response: Remapping vs. Deformation

Instead of deforming a single patch (which is complex and spatially expensive), ReloQate uses Remapping. The architecture reserves a set of "Reloqation Patches." When a tile's LER exceeds a target threshold, the logical qubit expands into an adjacent tile and contracts out of the old one—a process governed by Lattice Surgery.

Figure 2: A d=3 surface code undergoing a remap. The logical state is "moved" to a calibrated tile in constant time.

Experimental Results: Staying Below the Threshold

The authors tested ReloQate against two drift models: Slow Drift (hours-long degradation) and Volatile Drift (based on Google's Sycamore hardware data, featuring sudden jumps).

Key Findings:

• Predictor Accuracy: Larger DFR buffers provide smoother, more accurate LER estimates but can lag behind sudden "burst" errors. The optimal buffer size correlates with the drift speed.
• Spatial Efficiency: For code distances $d<11$, remapping requires fewer physical qubits than deformation-based methods because it doesn't require every qubit to have a dedicated "expansion zone."
• In-Situ Success: In memory experiments, ReloQate successfully kept logical qubits alive by hopping between tiles, maintaining a consistent LER even as the underlying hardware degraded.

Figure 3: LER history demonstrating that each a "remap" (the vertical drops) successfully resets the error accumulation.

Critical Analysis & Conclusion

ReloQate marks a significant shift from "Passive QEC" to "Active QEC Management."

Strengths:

• Zero Characterization Overhead: It uses data already being collected during syndrome extraction.
• Hardware Agnostic: While tested on Surface Codes, the DFR-LER logic applies to any stabilizer code (Color codes, qLDPC).

Limitations:

• Routing Congestion: Moving qubits frequently can "clog" the architecture, preventing other logical operations from executing (the so-called "cordoning off" effect).
• Scaling: As code distance $d$ increases, the relative benefit of remapping vs. deformation shifts, suggesting a hybrid approach is needed for future fault-tolerant machines.

Takeaway: Real-time DFR monitoring is the "Heart Rate Monitor" of a quantum computer. ReloQate provides the "Surgery" needed to keep the patient alive.