Small language models evaluated for smart data model classification on the edge (arXiv:2610.07093v1)
This arXiv preprint (2610.07093v1) benchmarks lightweight open-source language models — including general-purpose, reasoning-specialized, and code-specialized architectures — for classifying smart data models (SDMs) in resource-constrained edge/IoT settings, and compares them to near-zero-cost baselines (TF-IDF and a lightweight sentence encoder). The study reports model selection, task formulation, and deployment insights aimed at optimizing accuracy and efficiency for SDM resolution on edge platforms.