Recursive Code Evolution for Task-Oriented Semantic Communications
Abstract
Semantic communication jointly optimizes inference accuracy, channel use, and computation, yet the underlying communication program is typically specified before training. This paper presents ComRSI (Communications-Oriented Recursive Self-Improvement), a framework for recursively evolving that program through large-language-model-assisted code edits. A fixed agent modifies the encoder, channel adapters, decoder, loss, and training policy; a protected evaluator screens candidates under prescribed budgets and resource constraints. Accepted code and textual feedback inform subsequent proposals, while candidate weights are trained from scratch. In the original ten-generation CIFAR-100 lineage, eight accepted revisions produce an attention-pooled code achieving 74.3% test accuracy over an Additive White Gaussian Noise (AWGN) channel at 32 complex symbols per image, compared with 67.2% at 256 symbols for the initial design. The program also improves fading-channel accuracy and transfers to Tiny-ImageNet after retraining. A repeat study with ten searches per method compares ComRSI against non-recursive agents with and without textual history. Mean final search scores balancing accuracy, transmission rate, and computational cost are 0.7587, 0.7376, and 0.7089, respectively; the mean numbers of accepted revisions per search are 6.0, 5.0, and 2.1. The advantage over the memoryless agent is supported by a two-sided rank test, whereas the difference from the history-equipped agent remains statistically inconclusive at this sample size. Archived proposals show that pooling can yield a large improvement, while retained history supports subsequent refinement. These results support recursive program optimization within the evaluated setting and expose its remaining rate-robustness and compute-placement trade-offs.
Related
- AI-Native Task-Oriented Semantic Communications over a Relay: Benefits and Vulnerabilities
- Design of a Speech Enhancement-Based Deep Semantic Autoencoder for Wireless Communications
- Federated Learning for Semantic Communications over Multiple Access Channels
- Short-Length Code Designs for Integrated Sensing and Communications: A Deep Learning Approach
- More Interactions Make Communications More Efficient: A Directed Information Bottleneck Approach to Task-Oriented Communication
BibTeX
@misc{gong2026recursive,
title = {{Recursive Code Evolution for Task-Oriented Semantic Communications}},
author = {Yan Gong and Zheng Chu and Wei Koong Chai and Hari Mohan Pandey and Jingrui Hou and Miao Zhang and Yuanjian Li and Pei Xiao},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.29641},
eprint = {2610.29641},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.29641},
}
RIS
TY - UNPB
TI - Recursive Code Evolution for Task-Oriented Semantic Communications
AU - Yan Gong
AU - Zheng Chu
AU - Wei Koong Chai
AU - Hari Mohan Pandey
AU - Jingrui Hou
AU - Miao Zhang
AU - Yuanjian Li
AU - Pei Xiao
PY - 2026
DA - 2026/10/21
PB - arXiv
DO - 10.48550/arxiv.2610.29641
UR - https://arxiv.org/abs/2610.29641
AB - Semantic communication jointly optimizes inference accuracy, channel use, and computation, yet the underlying communication program is typically specified before training. This paper presents ComRSI (Communications-Oriented Recursive Self-Improvement), a framework for recursively evolving that program through large-language-model-assisted code edits. A fixed agent modifies the encoder, channel adapters, decoder, loss, and training policy; a protected evaluator screens candidates under prescribed budgets and resource constraints. Accepted code and textual feedback inform subsequent proposals, while candidate weights are trained from scratch. In the original ten-generation CIFAR-100 lineage, eight accepted revisions produce an attention-pooled code achieving 74.3% test accuracy over an Additive White Gaussian Noise (AWGN) channel at 32 complex symbols per image, compared with 67.2% at 256 symbols for the initial design. The program also improves fading-channel accuracy and transfers to Tiny-ImageNet after retraining. A repeat study with ten searches per method compares ComRSI against non-recursive agents with and without textual history. Mean final search scores balancing accuracy, transmission rate, and computational cost are 0.7587, 0.7376, and 0.7089, respectively; the mean numbers of accepted revisions per search are 6.0, 5.0, and 2.1. The advantage over the memoryless agent is supported by a two-sided rank test, whereas the difference from the history-equipped agent remains statistically inconclusive at this sample size. Archived proposals show that pooling can yield a large improvement, while retained history supports subsequent refinement. These results support recursive program optimization within the evaluated setting and expose its remaining rate-robustness and compute-placement trade-offs.
ER -