Structured Masked Diffusion for Joint Multiuser Decoding

2026-05-26

AI Summary

This work addresses the performance degradation and computational intractability of conventional joint decoding methods in high-load multi-user communication scenarios, where severe signal interference leads to ambiguous superimposed signals. To overcome these limitations, the authors propose CIDER—a learnable multi-user decoder grounded in a structured masking diffusion mechanism. CIDER enhances decoding robustness through a synergistic combination of interference cancellation, parity-aware message passing, and a confidence-guided dynamic remasking strategy. Its key innovation lies in a mask diffusion refinement process that prevents duplicate row collapse, complemented by a lightweight quality-guided redecoding module. Experimental results demonstrate that CIDER matches or surpasses the error-correction performance of FFT-accelerated joint belief propagation on standard codes, while achieving decoding speedups of 6× to over 100×, with gains becoming increasingly pronounced as code length grows.

Abstract

In joint multiuser decoding, a receiver recovers a set of messages from a single noisy aggregate of many simultaneous transmissions. Classical decoders rely on rule-based mechanisms such as successive interference cancellation, joint belief propagation, or list recovery, all of which become brittle or expensive as ambiguity increases. We propose CIDER, a learned multiuser decoder with masked-diffusion refinement steps. CIDER uses demixing to prevent duplicate-row collapse and uses parity-aware propagation to provide soft guidance from the code constraints. In higher-load regimes, we further improve reliability via a lightweight quality-guided remasking step that selectively re-decodes low-confidence sequences. On commonly used error-correcting codes, CIDER matches or improves on FFT-accelerated joint belief propagation-style decoding in symbol error rate while running more than $6\times$ to over $100\times$ faster, with the speedup widening as the blocklength grows. Code is available at https://github.com/jiyunyoung/CIDER.

Problem

• joint multiuser decoding
• noisy aggregate
• ambiguity
• symbol error rate
• error-correcting codes

Innovation

• masked diffusion
• multiuser decoding
• demixing
• parity-aware propagation
• quality-guided remasking