Riverlane's adaptive hardware decoder corrects errors in under 1 µs per round, cutting qubit overhead by up to 75%

Riverlane/Deltaflow

In a peer-reviewed Nature Communications paper, Riverlane described its Local Clustering Decoder, an FPGA decoder for the surface code that works in under a microsecond per round and adapts to leakage, where qubits slip out of their 0/1 states. Under leakage-heavy noise it reached target accuracy with a distance-17 code where a non-adaptive decoder needed distance 33.

• Latency: under 1 µs per decoding round.
• Adaptive decoding cuts physical-qubit overhead by up to 75%.
• Underpins Deltaflow, already deployed with Infleqtion, OQC, Rigetti and Oak Ridge National Laboratory.
• Deltaflow 3, with streaming logic, is targeted for late 2026.