Beyond Channel Data: Configuration-Enriched Radio Foundation Models

The paper introduces a method to enrich Radio Foundation Models (RFMs) by incorporating channel configuration (CFG) information during pre‑training. These CFG‑enriched embeddings are shown to act as effective inductive priors for downstream tasks, especially when labeled data are scarce. Experiments across three data regimes and multiple tasks—LoS/NLoS classification, timing advance estimation, and beam selection—demonstrate that the new embeddings consistently outperform or match baseline models in both in‑distribution and out‑of‑distribution configuration settings.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

• rag