arXiv:2606.00263v1 Announce Type: cross Abstract: 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 fol
The continuous evolution of wireless communication standards like 5G and the upcoming 6G necessitates more efficient and robust signal processing techniques, with AI/ML becoming central to these advancements.
Improved CSI denoising is crucial for enhancing the reliability and efficiency of MIMO-OFDM systems, leading to better spectral efficiency and reduced latency in a wide range of wireless applications.
This research introduces a length-generalizable AI model that can adapt to varying resource block allocations without retraining, significantly speeding up deployment and reducing computational overhead for wireless system optimization.
- · Telecommunications infrastructure providers
- · Mobile network operators
- · AI/ML researchers in wireless communication
- · Hardware manufacturers of 5G/6G devices
- · Traditional signal processing algorithm developers without AI integration
MIMO-OFDM systems will achieve higher throughput and greater reliability due to more accurate channel state information.
Enhanced wireless performance will accelerate the deployment and adoption of advanced 5G/6G applications requiring ultra-low latency and high data rates.
The reduced complexity of adapting AI models for different wireless configurations could foster more rapid innovation and deployment of next-generation communication technologies globally.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG