Abstract: In modern battlefields, rapid and accurate recognition of an adversary’s multifunction radar operating modes is of great importance for achieving battlefield superiority. However, in complex electromagnetic environments, signal overlap and interference often result in impure pulse description word (PDW) parameter sets, which significantly degrade the reliability of subsequent processing. To address this problem, this paper proposes a multifunction radar operating mode recognition method based on an Extended Long Short-Term Memory (xLSTM) network. First, the waveform characteristics of radar operating modes are analyzed, and a PDW-based dataset for multifunction radar signals is constructed. Then, leveraging the sequential feature extraction capability of the xLSTM network, an efficient and high-accuracy recognition algorithm for multifunction radar operating modes is developed. Through simulation experiments, the effectiveness of the sLSTM and mLSTM modules in the xLSTM architecture is validated. Furthermore, simulation tests are conducted under non-ideal conditions, including noise contamination, pulse loss, and spurious pulses, to evaluate the robustness of the proposed model. The results demonstrate that the proposed method achieves high accuracy and effectiveness in multifunction radar operating mode recognition.