analiza aiqbench
MultiPathFormer introduces a novel autoregressive foundation model for wireless communications by treating multipath propagation as the fundamental pretraining object rather than standard channel tensors. By leveraging environmental retrieval-augmented generation, this approach significantly improves path statistics estimation and demonstrates superior transferability across unseen users compared to existing state-of-the-art models. This work represents a significant step toward physically informed AI architectures that can generalize effectively in dynamic wireless environments without requiring extensive retraining from scratch.
MultiPathFormer: Towards a Foundation Model for Multipath Wireless Propagation
arXiv:2608.05076v1 Announce Type: new Abstract: Recent advances in machine learning have enabled training of wireless foundation models, which aim to support tasks such as channel estimation, beam prediction, and localization based on wireless signals. Existing wireless foundation models typically pretrain on channel tensors using masked reconstruction over subcarriers, antennas, or time but ignore the physical characteristics of wireless propagation. In this work, we propose to instead use multipath propagation as the fundamental pretraining object. We present MultiPathFormer, an autoregressive foundation model that represents each transmitter-receiver link as an ordered sequence of continuous-valued path tokens and pretrains with next-path prediction. We introduce an Environmental RAG (retrieval-augmented generation) mechanism and a first-path codebook on top of the transformer backbone, leveraging environment knowledge to improve path statistics estimation like delay and power by up to 59%. MultiPathFormer pretrained on 27 environments transfers to unseen users and, after scenario-specific fine-tuning, outperforms training the corresponding models from scratch in new environments. Across downstream tasks, it outperforms SOTA channel-based foundation models, achieving 5.57 m mean localization error, 0.914 top-3 beam accuracy, 0.994 line-of-sight classification accuracy, and 0.561 channel estimation NMSE. These results show that path-level pretraining can learn reusable representations of wireless propagation.
Abstract:Recent advances in machine learning have enabled training of wireless foundation models, which aim to support tasks such as channel estimation, beam prediction, and localization based on wireless signals. Existing wireless foundation models typically pretrain on channel tensors using masked reconstruction over subcarriers, antennas, or time but ignore the physical characteristics of wireless propagation. In this work, we propose to instead use multipath propagation as the fundamental pretraining object. We present MultiPathFormer, an autoregressive foundation model that represents each transmitter-receiver link as an ordered sequence of continuous-valued path tokens and pretrains with next-path prediction. We introduce an Environmental RAG (retrieval-augmented generation) mechanism and a first-path codebook on top of the transformer backbone, leveraging environment knowledge to improve path statistics estimation like delay and power by up to 59%. MultiPathFormer pretrained on 27 environments transfers to unseen users and, after scenario-specific fine-tuning, outperforms training the corresponding models from scratch in new environments. Across downstream tasks, it outperforms SOTA channel-based foundation models, achieving 5.57 m mean localization error, 0.914 top-3 beam accuracy, 0.994 line-of-sight classification accuracy, and 0.561 channel estimation NMSE. These results show that path-level pretraining can learn reusable representations of wireless propagation.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Signal Processing (eess.SP)
arXiv:2608.05076 [cs.LG]
(or arXiv:2608.05076v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2608.05076 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 292 cuvinte
Subiecte conexe
IEEE Celebrates Innovators Shaping the Future
IEEE Spectrum · 30 sept. 2026
From Next-Gen AI to Surgical Robots: IEEE Celebrates the Technical Innovators Shaping the Future
newswire.ca · 22 sept. 2026
News
Quantum Computing Report · 7 iun. 2026
A Comprehensive Survey of Wireless Foundation Models for AI-Native 6G Networks
arXiv · 18 aug. 2026