CSI-Agent: LLM-Assisted Few-Shot Adaptation for Cross-Domain Wi-Fi CSI Sensing

2026-09-25

AI Summary

This study addresses the class-specific performance degradation in cross-domain Wi-Fi CSI deployment caused by environmental or user variations, which existing methods struggle to mitigate effectively. We reformulate cross-domain adaptation as a deployment-time decision problem and propose an evidence-seeking LLM agent framework. By integrating multi-perspective CSI features to construct an adaptive default model, an LLM planner generates class-level evidence that undergoes deterministic verification to dynamically determine whether to retain default predictions or trigger specialized interventions. This approach enables fine-grained, verifiable few-shot cross-domain adaptation. Evaluated on four public datasets, the proposed method achieves state-of-the-art performance with only one-shot supervision, improving the average Macro-F1 score by approximately 16% over the strongest baseline.

Abstract

Wi-Fi channel state information (CSI) has enabled device-free sensing applications such as human activity recognition. However, CSI sensing models remain brittle in cross-domain deployment, where changes in users or environments can produce incorrect predictions. Existing solutions usually treat this problem as an offline model-design problem, by pretraining a stronger representation or applying one fixed adaptation method to the entire target domain. In practice, labeled target data are scarce and different classes may fail in different ways under the same domain shift. To address this, we propose CSI-Agent, an evidence-seeking LLM agent that reformulates cross-domain CSI adaptation as a deployment-time decision-making problem. Rather than processing raw CSI or making sample-level predictions, CSI-Agent summarizes target-domain behavior into sensing-grounded class-level evidence. It establishes a strong target-adaptive default from complementary CSI views and uses an LLM planner to determine whether each class should retain the default or invoke a specialized action. Deterministic verification and bounded execution further reduce unreliable interventions. We evaluate CSI-Agent on four public datasets using five cross-domain splits covering device, user, environment, and compositional shifts. Under 1-shot adaptation, CSI-Agent achieves the best target-domain performance across all splits and improves the average Macro-F1 by about 16\% compared to the strongest baseline method.

Problem

• Wi-Fi CSI sensing
• cross-domain adaptation
• few-shot learning
• human activity recognition
• domain shift

Innovation

• LLM Agent
• Cross-Domain Adaptation
• Few-Shot Learning
• Wi-Fi CSI Sensing
• Class-Level Evidence