Personal LLM Agents: Insights and Survey about the Capability, Efficiency and Security

code: https://github.com/MobileLLM/Personal_LLM_Agents_Survey

architecture, capibility, efficiency, security

capibility: task execution, context sensing, memorization.

efficiency: inference, customization, memory retrieval.

security & privacy: data confidentiality, decision reliability, system integrity.

Introduction

Personal LLM Agents v.s. Intelligent Personal Assistants

• flexibility and scalability
• IPAs: follow specific predefined rules to complete tasks, must explicitly specify which functions they wish to support, in addition to defining the triggers and steps for task execution.
• LLM Agents: solve complex problems by automatically using tools such as search engines, code interpreters, and third-party APIs.

Definition & Insights

Personal LLM Agents: a special type of LLM-based agent that is deeply integrated with personal data, personal devices, and personal services.

Intelligence Levels

Capibilities

Efficiency