Overcoming the BCI Calibration Bottleneck: A Clinically-Grounded Architecture using Riemannian Alignment and Stochastic Weight Averaging

Collection of 10 new arXiv papers (EE/ML/biomed/communications) covering: zero-calibration BCI using Riemannian alignment + stochastic weight averaging (high per-subject accuracy; hardware-agnostic LOSO performance); an LLM-driven "Agent Compiler" for translating high-level intent into executable ISAC (integrated sensing and communications) configurations with discussion of verification and pipeline security; physics-informed 1D-CNNs for multilayer cloud detection; efficient EEG seizure detection via INT8 quantization, channel pruning, and spiking neural networks (on-device optimization); QoS‑