Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Large Wireless Foundation Models: Stronger over Bigger

Xiang Cheng, Boxun Liu, Xuanyu Liu, Xuesong Cai

TL;DR

This paper introduces Large Wireless Foundation Models (LWFMs) as a framework to bring the generalization and reasoning strengths of foundation models into wireless systems while respecting stringent hardware and latency constraints. It outlines two complementary paradigms: (i) leveraging general-purpose foundation models through native multimodal capabilities or cross-modal transfer, and (ii) building wireless foundation models pretrained specifically on wireless data to yield universal representations and strong zero-shot capabilities. For each paradigm, the authors present representative roadmaps, design principles under wireless constraints, and key features, complemented by case studies (e.g., WiFo-2) demonstrating improved multi-tasking and adaptive communications with reduced data and parameter requirements. The discussion highlights current scale limitations, data, and multimodal integration challenges, and proposes a research agenda encompassing dataset development, multimodal information fusion, extended multitasking, efficient inference, and edge–device collaboration. Overall, LWFMs aim to deliver task-agnostic, robust wireless intelligence with practical deployment considerations, signaling a shift from task-specific models to versatile, scalable wireless AI systems.

Abstract

AI-communication integration is widely regarded as a core enabling technology for 6G. Most existing AI-based physical-layer designs rely on task-specific models that are separately tailored to individual modules, resulting in poor generalization. In contrast, communication systems are inherently general-purpose and should support broad applicability and robustness across diverse scenarios. Foundation models offer a promising solution through strong reasoning and generalization, yet wireless-system constraints hinder a direct transfer of large language model (LLM)-style success to the wireless domain. Therefore, we introduce the concept of large wireless foundation models (LWFMs) and present a novel framework for empowering the physical layer with foundation models under wireless constraints. Specifically, we propose two paradigms for realizing LWFMs, including leveraging existing general-purpose foundation models and building novel wireless foundation models. Based on recent progress, we distill two roadmaps for each paradigm and formulate design principles under wireless constraints. We further provide case studies of LWFM-empowered wireless systems to intuitively validate their advantages. Finally, we characterize the notion of "large" in LWFMs through a multidimensional analysis of existing work and outline promising directions for future research.

Large Wireless Foundation Models: Stronger over Bigger

TL;DR

This paper introduces Large Wireless Foundation Models (LWFMs) as a framework to bring the generalization and reasoning strengths of foundation models into wireless systems while respecting stringent hardware and latency constraints. It outlines two complementary paradigms: (i) leveraging general-purpose foundation models through native multimodal capabilities or cross-modal transfer, and (ii) building wireless foundation models pretrained specifically on wireless data to yield universal representations and strong zero-shot capabilities. For each paradigm, the authors present representative roadmaps, design principles under wireless constraints, and key features, complemented by case studies (e.g., WiFo-2) demonstrating improved multi-tasking and adaptive communications with reduced data and parameter requirements. The discussion highlights current scale limitations, data, and multimodal integration challenges, and proposes a research agenda encompassing dataset development, multimodal information fusion, extended multitasking, efficient inference, and edge–device collaboration. Overall, LWFMs aim to deliver task-agnostic, robust wireless intelligence with practical deployment considerations, signaling a shift from task-specific models to versatile, scalable wireless AI systems.

Abstract

AI-communication integration is widely regarded as a core enabling technology for 6G. Most existing AI-based physical-layer designs rely on task-specific models that are separately tailored to individual modules, resulting in poor generalization. In contrast, communication systems are inherently general-purpose and should support broad applicability and robustness across diverse scenarios. Foundation models offer a promising solution through strong reasoning and generalization, yet wireless-system constraints hinder a direct transfer of large language model (LLM)-style success to the wireless domain. Therefore, we introduce the concept of large wireless foundation models (LWFMs) and present a novel framework for empowering the physical layer with foundation models under wireless constraints. Specifically, we propose two paradigms for realizing LWFMs, including leveraging existing general-purpose foundation models and building novel wireless foundation models. Based on recent progress, we distill two roadmaps for each paradigm and formulate design principles under wireless constraints. We further provide case studies of LWFM-empowered wireless systems to intuitively validate their advantages. Finally, we characterize the notion of "large" in LWFMs through a multidimensional analysis of existing work and outline promising directions for future research.
Paper Structure (31 sections, 4 figures, 2 tables)

This paper contains 31 sections, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Large Wireless Foundation Models
  3. Motivation
  4. Challenges
  5. Paradigm 1: Leveraging General-purpose Foundation Models
  6. Concept
  7. Representative Roadmaps
  8. Harnessing intrinsic multimodal capabilities
  9. Cross-modal transfer of general knowledge
  10. Design Principles under Wireless Constraints
  11. Key Features
  12. Paradigm 2: Building Wireless Foundation Models
  13. Concept
  14. Representative Roadmaps
  15. Building universal wireless representations
  16. ...and 16 more sections

Figures (4)

  • Figure 1: An illustration of the proposed two paradigms for LWFM.
  • Figure 2: An illustration of the case study for LWFM-empowered AI-native communication system design. (a) illustrates the considered wireless multi-tasking scenario, where the BS serves multiple classes of users. (b) presents the simulation results of LWFM-enabled wireless multi-tasking. (c) presents the simulation results of LWFM-empowered intelligent adaptive communications.
  • Figure 3: An illustration of the proposed two paradigms for LWFM.
  • Figure 4: A comparison of the concept of “large” in LLMs and LWFMs.