published2025-2026
Distributionally Robust Wireless Semantic Communication with Large AI Models
Distributionally robust semantic transmission for large-model inference across noisy, bandwidth-limited, and shifting wireless environments.
JSAC 2026publication venue
Meaning over bits
WaSeCom studies what happens when the transmission target is not an exact bit sequence but the task-relevant meaning consumed by a large AI model. Wireless channels fluctuate, source distributions shift, and edge resources remain constrained; average-case training is not enough.
I contributed to modelling and experimental work for large-model inference under changing channels, including distributionally robust training simulations and evaluation in low-bandwidth settings. The resulting work appeared in the IEEE Journal on Selected Areas in Communications in 2026.