Cross-Attention Transformer for Joint Multi-Receiver Uplink Neural Decoding

Xavier Tardy 1,2 , Grégoire Lefebvre 2 , Apostolos Kountouris 2 , Haïfa Farès 1 , Amor Nafkha 1 · 2026

Read Cross-Attention Transformer for Joint Multi-Receiver Uplink Neural Decoding by Xavier Tardy 1,2 , Grégoire Lefebvre 2 , Apostolos Kountouris 2 , Haïfa Farès 1 , Amor Nafkha 1 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-04728.txt.

Description

Xavier Tardy and 4 co-authors design a neural receiver for Wi-Fi signals gathered cooperatively by several access points in a 2026 paper. Pooling those receivers could improve uplink decoding, but fusing their signals well is difficult.

A shared encoder first learns the time-frequency structure at each receiver separately. A cross-attention module then fuses the receivers token by token, producing soft reliability scores for a standard channel decoder. No explicit channel estimates are needed. Training used a bit-metric objective, and the fusion adapts to each receiver's reliability. Over realistic Wi-Fi channels, the model beat classical pipelines and strong neural baselines. It often matched a reference with perfect channel knowledge while staying compact.

Catalog record

CREATORXavier Tardy 1,2 , Grégoire Lefebvre 2 , Apostolos Kountouris 2 , Haïfa Farès 1 , Amor Nafkha 1
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-04728

Browse the full library · About the Tower of Babel