Abstract: A converged radio unit (RU) configured for semantic-less retransmissions comprises a radio-frequency machine learning (RFML) radio unit (RFML RU) and a radio-access network (RAN) radio unit (RU). The RFML RU may be configured to receive an RF signal comprising a packet, decode a PHY layer preamble of the packet, and apply a machine-learning (ML) model to information obtained from only the PHY layer preamble to classify the packet. When the packet is classified be a tactical data link (TDL) packet based on the information obtained only from the PHY layer preamble, the RFML RU may retransmit the packet. When the packet is classified a RAN protocol packet, the RFML RU may be configured to signal the RAN RU to process the RAN protocol packet in accordance with a RAN protocol.