Towards Native AI in 6G Standardization: The Roadmap of Semantic Communication

TL;DR

This paper presents a comprehensive overview of recent progress in SemCom from both academic and industrial perspectives, with a focus on its ongoing and upcoming standardization activities.

Abstract

Semantic communication (SemCom) has emerged as a transformative paradigm for future 6G networks, offering task-oriented and meaning-aware transmission that fundamentally redefines traditional bit-centric design. Recognized by leading standardization bodies including the institute of electrical and electronics engineers (IEEE) and the international telecommunication union (ITU), and actively discussed within the 3rd generation partnership project (3GPP) working groups, SemCom is rapidly gaining traction as a foundational enabler for native-AI 6G. This paper presents a comprehensive overview of recent progress in SemCom from both academic and industrial perspectives, with a focus on its ongoing and upcoming standardization activities. We systematically examine advances in representative application scenarios, architectural design, semantic-traditional system compatibility, unified evaluation metrics, and validation methodologies. Furthermore, we highlight several key enabling technologies, such as joint source-channel coding (JSCC), SemCom-based multiple access (MA) technologies such as model division MA (MDMA), and semantic knowledge base (KB), that support the practical implementation of SemCom in standard-compliant systems. Additionally, we present a case study for channel state information (CSI) feedback, illustrating the concrete performance gains of SemCom under 3GPP-compliant fading channels. Finally, we discuss emerging challenges and research opportunities for incorporating semantic-native mechanisms into the evolving 6G standardization landscape, and provide forward-looking insights into its development and global adoption.

Figures (4)

• Figure 1: Typical scenarios and E2E architecture of 6G SemCom. The upper part of the illustration shows four emblematic use cases, i.e., satellite communications, immersive communication, intelligent internet of vehicles, and multi-AI agent collaboration. The lower part depicts a SemCom architecture that underpins them. By supporting these scenarios, the SemCom architecture enables optimized spectrum efficiency, reduced end-to-end latency, and context-adaptive services across heterogeneous network layers and application domains.
• Figure 2: Illustrative framework of key enabling SemCom technologies for 6G. The diagram visualizes the end‑to‑end data flow, functional interplay, and evolution pathway of four core technologies, i.e., A: JSCC, B: SemCom-based MA, C: SemCom‑based CSI Feedback, and D: Semantic KB Technologies.
• Figure 3: Compatibility design framework for semantic-traditional communication systems. The framework overlays new semantic-aware blocks onto the traditional mobile-network protocol stack to support seamless deployment of SemCom alongside existing procedures such as CRC, HARQ, and resource scheduling.
• Figure 4: Squared generalized cosine similarity (SGCS) vs. SNR for the CSI feedback case study. "JSCCM" denotes the joint source-channel coding and modulation scheme. "JSCC+QPSK" applies deep JSCC for compression and quadrature phase shift keying (QPSK) for modulation. The remaining schemes represent SSCC: "32bit+1/4LDPC+QPSK" (and similar variants) indicates that an AI model performs source coding, producing a 32-bit representation that is then protected by a rate-1/4 LDPC code and modulated with QPSK. "JSCC+QPSK" and all SSCC baselines use 2-bit quantization so that each scheme transmits 128 symbols over the air.