Recently, tech enthusiasts have observed significant performance degradation in Google Gemini and OpenAI ChatGPT after user authentication, manifesting as reduced answer quality, shortened context, and simplified implementations. The author analyzes that this may be due to platform-side prompt strategies limiting the models’ deep thinking capabilities. To address this issue, the author shares a meticulously designed prompt set covering scenarios such as paper analysis, academic advice expert, and illustration expert. Through detailed few-shot examples and operational guidelines, it demonstrates how to optimize prompts to enhance AI output quality in academic paper interpretation, code model improvement, and scientific illustration. These practical techniques not only reveal the limitations of AI models but also provide users with effective performance optimization solutions, offering high reference value for readers focused on cutting-edge AI technology.
