Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Leveraging Large Language Models for Wireless Symbol Detection via In-Context Learning

Momin Abbas, Koushik Kar, Tianyi Chen

TL;DR

This work tackles wireless symbol demodulation under severe data scarcity, where traditional DNNs underperform. It introduces LMIC, a no-training approach that exploits in-context learning in large language models, augmented with calibration methods ConC and LinC to stabilize and quantify reliability. Across experiments with $K=8$ APSK at $SNR = 5~\mathrm{dB}$, LMIC variants frequently surpass DNN baselines, demonstrating higher accuracy and better-calibrated outputs (lower entropy and ECE), particularly for larger models like Llama-2 $7$B and $13$B in 32-shot settings. The results indicate that no-training LLM-based inference, when properly calibrated, can be a competitive and reliable alternative for non-language wireless tasks, with potential for no-code deployment in data-constrained environments.

Abstract

Deep neural networks (DNNs) have made significant strides in tackling challenging tasks in wireless systems, especially when an accurate wireless model is not available. However, when available data is limited, traditional DNNs often yield subpar results due to underfitting. At the same time, large language models (LLMs) exemplified by GPT-3, have remarkably showcased their capabilities across a broad range of natural language processing tasks. But whether and how LLMs can benefit challenging non-language tasks in wireless systems is unexplored. In this work, we propose to leverage the in-context learning ability (a.k.a. prompting) of LLMs to solve wireless tasks in the low data regime without any training or fine-tuning, unlike DNNs which require training. We further demonstrate that the performance of LLMs varies significantly when employed with different prompt templates. To solve this issue, we employ the latest LLM calibration methods. Our results reveal that using LLMs via ICL methods generally outperforms traditional DNNs on the symbol demodulation task and yields highly confident predictions when coupled with calibration techniques.

Leveraging Large Language Models for Wireless Symbol Detection via In-Context Learning

TL;DR

This work tackles wireless symbol demodulation under severe data scarcity, where traditional DNNs underperform. It introduces LMIC, a no-training approach that exploits in-context learning in large language models, augmented with calibration methods ConC and LinC to stabilize and quantify reliability. Across experiments with APSK at , LMIC variants frequently surpass DNN baselines, demonstrating higher accuracy and better-calibrated outputs (lower entropy and ECE), particularly for larger models like Llama-2 B and B in 32-shot settings. The results indicate that no-training LLM-based inference, when properly calibrated, can be a competitive and reliable alternative for non-language wireless tasks, with potential for no-code deployment in data-constrained environments.

Abstract

Deep neural networks (DNNs) have made significant strides in tackling challenging tasks in wireless systems, especially when an accurate wireless model is not available. However, when available data is limited, traditional DNNs often yield subpar results due to underfitting. At the same time, large language models (LLMs) exemplified by GPT-3, have remarkably showcased their capabilities across a broad range of natural language processing tasks. But whether and how LLMs can benefit challenging non-language tasks in wireless systems is unexplored. In this work, we propose to leverage the in-context learning ability (a.k.a. prompting) of LLMs to solve wireless tasks in the low data regime without any training or fine-tuning, unlike DNNs which require training. We further demonstrate that the performance of LLMs varies significantly when employed with different prompt templates. To solve this issue, we employ the latest LLM calibration methods. Our results reveal that using LLMs via ICL methods generally outperforms traditional DNNs on the symbol demodulation task and yields highly confident predictions when coupled with calibration techniques.
Paper Structure (9 sections, 7 equations, 4 figures, 4 tables)

This paper contains 9 sections, 7 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Context and Motivation
  3. Related Work
  4. Formulation and Solution Approach
  5. Wireless Symbol Demodulation
  6. Deep Neural Networks (DNNs)
  7. Proposed Approach: LLM-based ICL (LMIC)
  8. Numerical Evaluation
  9. Conclusions

Figures (4)

  • Figure 1: Example of ICL with LLM $\theta^*$.
  • Figure 2: Comparison across ten different prompt templates.
  • Figure 3: Shannon entropy histograms of different ICL methods on Llama-2 13B for 4/8-shots setting; we use logarithmic base two.
  • Figure 4: Shannon entropy histograms of different ICL methods on Llama-2 13B for 16/32-shot setting; we use logarithmic base two.