TONIC: Token-Centric Semantic Communication for Task-Oriented Wireless Systems
This work addresses the mismatch between bit-level wireless transmission and token-level task relevance for foundation models in task-oriented communication, offering a modular and interpretable alternative to black-box end-to-end learning.
TONIC proposes a token-centric semantic communication framework for task-oriented wireless systems that allocates unequal error protection based on token-level task relevance and uses confidence-aware gating at the receiver to replace harmful substitutions with erasures, which are then restored by a Transformer-based completion model. On image classification tasks, TONIC consistently outperforms separation-based schemes, DeepJSCC, and token-domain baselines under matched communication budgets across AWGN, Rayleigh, and Rician channels.
Tokens are becoming the basic units through which foundation models represent and process information for understanding and inference. However, traditional wireless communication, centered on bit-level fidelity, faces a mismatch between what is transmitted reliably and what downstream models actually consume. This mismatch calls for a communication design that directly accounts for token-level task relevance and downstream model requirements, rather than treating all transmitted bits as equally important. In this paper, we propose TONIC, a token-centric semantic communication framework for task-oriented wireless systems. The transmitter converts each source sample into a sequence of tokens, estimates token-level task relevance, and allocates protection through utility-aware unequal error protection under a fixed channel-use budget. At the receiver, token-level confidence is used to gate unreliable decisions, turning harmful substitutions into recoverable erasures before a Transformer-based completion model restores the masked tokens for final task inference. Our framework combines transmitter-side semantic-aware protection with receiver-side confidence-aware gating in a modular and interpretable architecture, rather than relying solely on fully black-box end-to-end learning. We further establish a utility-aware Bayes-risk interpretation for the receiver-side gating rule and study its interaction with unequal protection and completion. Experimental results on image classification show that TONIC consistently outperforms separation-based schemes, the pixel-domain DeepJSCC baseline, and token-domain baselines under matched communication budgets over AWGN, Rayleigh, and Rician channels.