The field of large language models (LLMs) is moving towards personalized models that can adapt to individual user preferences and needs. This is motivated by the need to improve user experience and engagement in various applications such as conversational AI, online learning, and web navigation. Researchers are exploring techniques to induce synthetic user personas, develop personalized reward models, and create benchmark datasets to evaluate the effectiveness of these models. Notably, there is a growing interest in developing agentic web interfaces that are optimized for agents rather than humans, which could potentially overcome the limitations of current web agent designs.

Some noteworthy papers in this area include: PersonaAgent, which introduces a personalized LLM agent framework that integrates episodic and semantic memory mechanisms to respond to users' varying needs and preferences. OPeRA, which presents a novel dataset for evaluating LLMs on human online shopping behavior simulation, capturing user personas, browser observations, fine-grained web actions, and self-reported rationales. Build the web for agents, not agents for the web, which advocates for a paradigm shift in web agent research, introducing the concept of an Agentic Web Interface (AWI) specifically designed for agents to navigate a website.