Abstract:
The rapid evolution of large language models (LLMs) has opened new avenues for enhancing healthcare delivery, particularly through cloud-edge collaborative frameworks. Th...Show MoreMetadata
Abstract:
The rapid evolution of large language models (LLMs) has opened new avenues for enhancing healthcare delivery, particularly through cloud-edge collaborative frameworks. This paper introduces PrismPrompt, a novel system that leverages prompt-based engineering to optimize cloud-edge collaboration in medical applications. By integrating cloud-based LLMs with edge devices, PrismPrompt addresses the challenges of computational limitations and data privacy in healthcare environments. The system utilizes a hierarchical prompt strategy and an incremental expert decision-making process to enhance the retrieval and application of medical knowledge. Key innovations include a retriever module that accurately extracts and retrieves relevant information from cloud models and a decision maker that synthesizes expert opinions to ensure accurate and context-aware medical advice. Experimental results demonstrate that PrismPrompt outperforms existing models in terms of accuracy, highlighting its potential to improve real-time medical decision-making while preserving the computational feasibility on edge devices. This work provides a promising step towards the broader adoption of cloud-edge collaborative LLMs in healthcare, offering scalable and privacy-conscious solutions for modern medical challenges.
Published in: IEEE Network ( Early Access )