Real-Time Cross-Layer Semantic Error Correction Using Language Models and Software-Defined Radio

TL;DR

This paper demonstrates a real-time implementation of Cross-Layer Semantic Error Correction using language models and SDR hardware, significantly improving accuracy over previous methods.

Contribution

It validates the real-time feasibility of CL-SEC on SDR hardware with new middleware and inference interfaces, enabling practical deployment.

Findings

Cross-layer fusion outperforms individual sources in error correction

Real-time LLR extraction from FPGA hardware is achieved

The system improves semantic error correction accuracy in live tests

Abstract

As Language Models (LMs) advance, Semantic Error Correction (SEC) has emerged as a promising approach for reliable network designs. Yet existing methods prioritize intent over accuracy, falling short of verbatim recovery. Our recent work, Cross-Layer SEC (CL-SEC), addressed this by fusing physical-layer Log-Likelihood Ratios (LLRs) with semantic context, but its real-time feasibility remained unvalidated. This paper demonstrates CL-SEC on a live Software-Defined Radio (SDR) testbed, resolving implementation barriers with: 1) an SDR middleware enabling real-time LLR extraction from FPGA hardware, and 2) a generalized inference interface supporting modern encoder-decoder LMs. Real-world experiments confirm that the cross-layer fusion significantly outperforms either source alone.