Researchers propose MANE, a multi-path adaptive network designed to improve split computing, where a small head model runs on a device and a larger tail model runs on an edge server. The approach targets efficient distributed inference by adapting how computation is divided between the device and the edge. It is described in a new arXiv preprint (2609.14660v1) listed under cross-submissions.
Researchers introduced FedLTLib, a benchmark suite for federated learning in settings where data across clients follows a long-tailed distribution. The work addresses real-world mobile and edge deployments, where privacy constraints keep data decentralized and class frequencies are highly uneven. The benchmark aims to standardize evaluation of methods designed for this combination of challenges.
A new arXiv paper introduces FREDI, a security-focused wireless edge-intelligence framework for event-triggered inference across user devices, edge servers, and the cloud. It combines proportional-fair resource allocation with a dual-threshold early-exit scheme so that each user device can partially process inference locally before offloading. The work aims to balance fairness, latency, and efficiency in cooperative multi-layer edge deployments.
Researchers developed a machine learning approach that forecasts railway delays in Finland by combining weather observations with other operational data. The work situates such environmental sensing within future 6G-enabled wireless sensor networks and edge computing, which the authors argue will improve real-time reliability. The study is published as an arXiv preprint.
A new arXiv paper introduces mmFHE, a system that performs an entire cloud-side mmWave sensing workflow, including signal processing and machine learning inference, on encrypted data using fully homomorphic encryption. Range profiles are encrypted on the edge device first, so the cloud never sees raw sensing data. The work aims to enable privacy-preserving sensing services without giving up cloud compute.
Researchers propose BiHDTrans, a transformer variant that uses binary hyperdimensional computing to classify multivariate time series from IoT sensors. The design targets resource-constrained edge devices, where large data volumes and limited compute make standard models impractical. It is presented as an arXiv preprint focused on balancing efficiency with classification accuracy.
A new research paper proposes a federated learning framework that separates modality-specific processing so heterogeneous robots in 6G networks can train collaboratively without sharing raw sensor data. The approach targets privacy preservation for embodied AI applications built on low-latency edge connectivity and distributed sensing.
A new arXiv paper surveys how large language models can support deep reinforcement learning in managing resources across IoT, edge, and cloud layers. The work focuses on continuous, context-aware decision-making in environments where constraints shift constantly. It positions LLMs as a complement to established DRL techniques for adaptive computing across the computing continuum.