Large and Small Model Collaboration for Air Interface

2025-12-12

AI Summary

Large AI Models (LAMs) struggle to simultaneously achieve strong generalization and environment-specific adaptation in wireless communications, while fine-tuning them incurs prohibitive computational cost, catastrophic forgetting, and parameter inaccessibility. Method: This paper proposes LAM-SAM—a collaborative air-interface framework—where a LAM serves as a universal channel knowledge base, and lightweight Small Adaptation Modules (SAMs) act as plug-and-play environment adapters. We introduce LASCO and its elastic variant E-LASCO, featuring a reference-agent dual-SAM architecture and learnable collaboration coefficients to dynamically balance generalization and scenario-specific adaptation. Contribution/Results: Experiments demonstrate that LAM-SAM significantly reduces training overhead and data requirements, accelerates environmental adaptation, and achieves substantial performance gains in CSI feedback tasks. Moreover, its inference latency meets real-time multi-user communication constraints, enabling practical deployment.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Multiagent Systems: Agent Communication
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Large artificial intelligence models (LAMs) have shown strong capability in wireless communications, yet existing works mainly rely on their generalized knowledge across environments while overlooking the potential gains of environment-specific adaptation. Directly fine-tuning LAMs for adaptation is often impractical due to prohibitive training costs, low inference efficiency in multi-user scenarios, and the risk of catastrophic forgetting, in addition to the limited accessibility of model parameters. To address these limitations, we establish a collaborative framework for air interface. In this framework, unlike prior approaches that either depend solely on LAMs or require direct fine-tuning, LAMs are exploited as a universal channel knowledge base while small artificial intelligence models (SAMs) are employed as lightweight plugins to capture environment-specific knowledge, facilitating efficient environment-specific adaptation of LAMs. Subsequently, we instantiate this framework for CSI feedback tasks, and develop a large and small collaboration framework for CSI feedback, referred to as LASCO. LASCO operates by letting the base LAM produce an initial CSI reconstruction, learning the environment-induced reconstruction shift through a reference SAM and a proxy SAM, and transferring this shift back to the LAM. To further enhance adaptability, we introduce elastic-LASCO (E-LASCO), which augments LASCO with learnable collaboration coefficients that control the contribution of LAMs and SAMs across different environments. Numerical results demonstrate that LASCO and E-LASCO enables LAMs to achieve environment-specific performance gains with significantly reduced training costs, lower data collection requirements, and faster adaptation speed.

Problem

• Enabling large AI models to adapt efficiently to specific wireless environments.
• Reducing training costs and data needs for environment-specific model adaptation.
• Preventing catastrophic forgetting while improving inference efficiency in multi-user scenarios.

Innovation

• Collaborative framework uses large models as universal knowledge base
• Small models act as lightweight plugins for environment-specific adaptation
• Elastic coefficients optimize contributions across different environments dynamically