The research area is witnessing a significant shift towards more efficient and versatile methods for prompt optimization and user targeting, driven by advancements in large language models (LLMs). A notable trend is the integration of gradient-based optimization techniques, which allow for more precise and computationally efficient prompt refinement, even in smaller models. This approach not only reduces dependency on massive LLMs but also enhances the transferability and performance of optimized prompts across various tasks and domains. Additionally, there is a growing emphasis on interactive and multi-objective optimization frameworks that incorporate human input and diverse criteria, respectively, to tailor prompts more effectively to specific contexts and user needs. These developments are paving the way for more robust and adaptable systems in fields such as digital marketing and recommendation systems, where cross-domain transferability and real-time forecastability are critical. Notably, innovative models like FIND are demonstrating superior performance in industrial-grade user targeting, while systems like iPrOp and MOPO are setting new standards for interactive and multi-objective prompt optimization, respectively. These advancements collectively underscore a move towards more intelligent, human-in-the-loop, and context-aware AI systems.