JPAF applies Carl Jung’s psychological theory to give LLMs structured, adaptive personalities via three mechanisms: dominant-auxiliary coordination for core consistency, reinforcement-compensation for contextual adaptation, and reflection for long-term evolution. The framework supports MBTI questionnaire evaluation and type-specific scenario testing. It has been validated on mainstream models including GPT, Llama, and Qwen, targeting applications in HCI, personalized assistants, and social simulations.