ReFLEX: Length-Generalizable CSI Denoising for MIMO-OFDM via Relative-Frequency Bias
This work addresses the need for flexible CSI denoising in 5G NR systems with variable resource block allocations, enabling a single model to handle unseen lengths without retraining.
ReFLEX introduces a length-generalizable Transformer for CSI denoising in MIMO-OFDM, achieving about -9.6 dB NMSE on unseen RB lengths and reducing the 10% BLER threshold by 2-3 dB in NR PUSCH/UL-SCH simulations.
This letter studies CSI denoising for MIMO--OFDM with variable NR resource block (RB) allocations. ReFLEX is a length-generalizable Transformer whose frequency attention uses a relative-frequency position bias (RFPB) generated from subcarrier offsets. A single checkpoint handles unseen RB lengths and can be applied to sparse DM-RS observations in the tested RB5/RB10 PUSCH setup without retraining. In a 3GPP~TR~38.901 UMa NLOS channel, ReFLEX achieves about $-9.6$~dB NMSE on unseen RB lengths. In NR PUSCH/UL-SCH simulations, ReFLEX denoising followed by time-frequency interpolation reduces the 10\% BLER threshold by about 2--3~dB.