Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Plug-and-Play AMC: Context Is King in Training-Free, Open-Set Modulation with LLMs

Mohammad Rostami, Atik Faysal, Reihaneh Gh. Roshan, Huaxia Wang, Nikhil Muralidhar, Yu-Dong Yao

TL;DR

This work tackles automatic modulation classification (AMC) under challenging, noisy wireless environments by proposing a training-free, in-context approach that leverages higher-order statistics and cumulants. The method converts quantitative signal features into structured natural-language prompts and employs exemplar-context prompts to guide a pretrained large language model in one-shot modulation identification, including open-set scenarios. Experimental results on synthetic data show that incorporating exemplar context and using larger models significantly improves accuracy across modulation schemes and SNRs, with the o3-mini 200B model achieving up to ~69.9% overall and ~72.4% cleaned accuracy under noise. The study demonstrates a promising direction for foundation-model-based wireless signal processing, reducing the need for channel-specific training while maintaining robustness and interpretability.

Abstract

Automatic Modulation Classification (AMC) is critical for efficient spectrum management and robust wireless communications. However, AMC remains challenging due to the complex interplay of signal interference and noise. In this work, we propose an innovative framework that integrates traditional signal processing techniques with Large-Language Models (LLMs) to address AMC. Our approach leverages higher-order statistics and cumulant estimation to convert quantitative signal features into structured natural language prompts. By incorporating exemplar contexts into these prompts, our method exploits the LLM's inherent familiarity with classical signal processing, enabling effective one-shot classification without additional training or preprocessing (e.g., denoising). Experimental evaluations on synthetically generated datasets, spanning both noiseless and noisy conditions, demonstrate that our framework achieves competitive performance across diverse modulation schemes and Signal-to-Noise Ratios (SNRs). Moreover, our approach paves the way for robust foundation models in wireless communications across varying channel conditions, significantly reducing the expense associated with developing channel-specific models. This work lays the foundation for scalable, interpretable, and versatile signal classification systems in next-generation wireless networks. The source code is available at https://github.com/RU-SIT/context-is-king

Plug-and-Play AMC: Context Is King in Training-Free, Open-Set Modulation with LLMs

TL;DR

This work tackles automatic modulation classification (AMC) under challenging, noisy wireless environments by proposing a training-free, in-context approach that leverages higher-order statistics and cumulants. The method converts quantitative signal features into structured natural-language prompts and employs exemplar-context prompts to guide a pretrained large language model in one-shot modulation identification, including open-set scenarios. Experimental results on synthetic data show that incorporating exemplar context and using larger models significantly improves accuracy across modulation schemes and SNRs, with the o3-mini 200B model achieving up to ~69.9% overall and ~72.4% cleaned accuracy under noise. The study demonstrates a promising direction for foundation-model-based wireless signal processing, reducing the need for channel-specific training while maintaining robustness and interpretability.

Abstract

Automatic Modulation Classification (AMC) is critical for efficient spectrum management and robust wireless communications. However, AMC remains challenging due to the complex interplay of signal interference and noise. In this work, we propose an innovative framework that integrates traditional signal processing techniques with Large-Language Models (LLMs) to address AMC. Our approach leverages higher-order statistics and cumulant estimation to convert quantitative signal features into structured natural language prompts. By incorporating exemplar contexts into these prompts, our method exploits the LLM's inherent familiarity with classical signal processing, enabling effective one-shot classification without additional training or preprocessing (e.g., denoising). Experimental evaluations on synthetically generated datasets, spanning both noiseless and noisy conditions, demonstrate that our framework achieves competitive performance across diverse modulation schemes and Signal-to-Noise Ratios (SNRs). Moreover, our approach paves the way for robust foundation models in wireless communications across varying channel conditions, significantly reducing the expense associated with developing channel-specific models. This work lays the foundation for scalable, interpretable, and versatile signal classification systems in next-generation wireless networks. The source code is available at https://github.com/RU-SIT/context-is-king
Paper Structure (13 sections, 8 equations, 2 figures, 2 tables, 1 algorithm)

This paper contains 13 sections, 8 equations, 2 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Theoretical Background
  3. High-Order Statistics
  4. Method
  5. Signal Summarization
  6. Example Context Construction
  7. Prompt Formulation
  8. Generation and Post-Processing
  9. Experiments and Results
  10. Dataset Generation
  11. Experimental Setup
  12. Overall Classification Performance
  13. Conclusion

Figures (2)

  • Figure 1: Overview of our in-context AMC framework. The raw signal is first summarized into high-order statistics and cumulants, which are then embedded in a text prompt. This prompt is fed into a large language model to perform modulation classification from a predefined set of modulation types (e.g., 4ASK, 4PAM, 8ASK, 16PAM).
  • Figure 2: Confusion matrix for DeepSeek‑R1‑Distill‑Qwen‑32B using the I+C+S prompt on noisy signals.