Learning to Decode Quantum LDPC Codes via Cluster-Based Sequential Belief Propagation

arXiv:2607.20130 2026 Dynamics 2 ideas extracted · analyzed Aug 30, 2026

What the math gives to ML

The transferable contribution is a residual-aware controller for scheduling local updates in sparse iterative computations. It combines synchronous updates within independently selected clusters with a permutation-invariant, quantized histogram of local mismatch statistics, so the controller state does not grow with cluster size. This structure can be transplanted into recurrent GNNs and learned iterative inference networks, where updating every node at every micro-step is expensive. A secondary application is selective block optimization for parameter-efficient fine-tuning, although that transfer is less direct.

Ideas from this paper

Mechanism confirmed, baseline not beaten 2026

Histogram-Controlled Cluster Updates for Iterative GNNs

Replace node-by-node scheduling in an iterative message-passing network with a learned scheduler that selects one graph cluster at a time, while updating all nodes in that cluster synchronously. The scheduler observes a quantized histogram of local residual weights, making its state invariant to permutations of nodes inside a cluster and independent of cluster cardinality.

Useful7/10
Difficulty5/10
Novelty7/10
Paper: Learning to Decode Quantum LDPC Codes via Cluster-Based Sequential Belief Propagation arXiv:2607.20130
Unverified 2026

Residual-Histogram Block Coordinate Fine-Tuning

Use the cluster-state construction to schedule which groups of trainable parameters receive an expensive update at each optimizer micro-step. Instead of updating every LoRA block, expert group, or layer uniformly, select the block whose local error histogram predicts the largest loss reduction per unit compute.

Useful5/10
Difficulty6/10
Novelty6/10
Paper: Learning to Decode Quantum LDPC Codes via Cluster-Based Sequential Belief Propagation arXiv:2607.20130