arxivcs.AI2026-07-16
SmartRAG: Native Graph-Based RAG for Mobile Device
Zhihan Jiang, Meng Li, Shenghao Liu, Keran Li, Ruiben Zhou, Wei Wang, et al.
Deploying large language models (LLMs) as personal assistants on mobile devices demands privacy, low latency, and offline availability, yet the computational cost of giant models clashes with strict edge-hardware budgets. We argue that this tension cannot be resolved by model com…