Title:NEFT: A Unified Transformer Framework for Efficient Near-Field CSI Feedback in XL-MIMO Systems

Abstract:Extremely large-scale multiple-input multiple-output (XL-MIMO) systems, operating in the near-field region due to their massive antenna arrays, are a key enabler of next-generation wireless communications but face significant challenges in channel state information (CSI) feedback. Deep learning has emerged as a powerful tool by learning compact CSI representations for feedback. However, existing methods struggle to capture the intricate structure of near-field CSI while incurring prohibitive computational overhead on practical mobile devices. To overcome these limitations, we propose the Near-Field Efficient Feedback Transformer (NEFT) family for accurate and efficient near-field CSI feedback across diverse hardware platforms. Built on a hierarchical Vision Transformer backbone, NEFT is extended with lightweight variants to meet various deployment constraints: NEFT-Compact applies multi-level knowledge distillation (KD) to reduce complexity while maintaining accuracy, and NEFT-Hybrid and NEFT-Edge address encoder- and edge-constrained scenarios via attention-free encoding and KD. Extensive simulations show that NEFT achieves a 15--21 dB improvement in normalized mean-squared error (NMSE) over state-of-the-art methods, while NEFT-Compact and NEFT-Edge reduce total FLOPs by 25--36% with negligible accuracy loss. Moreover, NEFT-Hybrid lowers encoder-side complexity by up to 64%, enabling deployment in highly asymmetric device scenarios. These results establish NEFT as a practical and scalable solution for near-field CSI feedback in XL-MIMO systems.