Plaquette v2025.4 released with loss-aware decoding support

We are excited to announce Plaquette v2025.4, introducing loss-aware decoding—a powerful technique that boosts the performance of fault-tolerant quantum computers.  Loss-aware decoding leverages information about which qubits have been lost during computation and where, thus enabling quantum computers to tolerate higher qubit losses.

The problem of losses

Qubit losses are quite hard to deal with: they are present in some form or another across different hardware platforms. And they can only be detected by specialized measurements, which we call loss-detection units. Furthermore, losses can occur far away from their detection point, so their presence can impact the neighbouring qubits and disrupt the computation before the detection happens. While traditional decoders can partially take into account that a specific qubit was lost, they do not consider how computation could have been affected depending on where the loss occurred.

Our new loss-aware decoding features improve decoding performance by taking into account all the different locations where a qubit could have been lost and the related impact on the quantum computation. In this way, the users can obtain more accurate decoder settings that improve their decoding performance, reducing the overall logical error rates.

With loss-aware decoding now in Plaquette, architecture and hardware teams can accurately benchmark quantum error-correcting codes under realistic loss conditions, tailor decoders to device-specific erasure profiles, and accelerate the development of fault-tolerant quantum architectures. 

To benchmark the performance of loss-aware decoding, we show that logical error rates improve as we add more loss detection units in the circuits: