PersonaAgent: When Large Language Model Agents Meet Personalization at Test Time
This limitation motivates us to develop PersonaAgent, the first personalized LLM agent framework designed to address versatile personalization tasks. Specifically, PersonaAgent integrates two complementary components - a personalized memory module that includes episodic and semantic memory mechanisms; a personalized action module that enables the agent to perform tool actions tailored to the user. At the core, the persona (defined as unique system prompt for each user) functions as an intermediary: it leverages insights from personalized memory to control agent actions, while the outcomes of these actions in turn refine the memory. Based on the framework, we propose a test-time user-preference alignment strategy that simulate the latest n interactions to optimize the persona prompt, ensuring real-time user preference alignment through textual loss feedback between simulated and ground-truth responses. Experimental evaluations demonstrate that PersonaAgent significantly outperforms other baseline methods by not only personalizing the action space effectively but also scaling during test-time real-world applications. These results underscore the feasibility and potential of our approach in delivering tailored, dynamic user experiences.
As Table 1 highlights, achieving effective personalization intelligence can be measured from four critical perspectives: agentic intelligence, real-world applicability, personal data utilization, and preference alignment. Yet, balancing these dimensions simultaneously remains a fundamental challenge. Early efforts for aligning LLMs with human preferences, such as supervised fine-tuning (Zhang et al., 2023) and reinforcement learning from human feedback (RLHF) (Schulman et al., 2017; Rafailov et al., 2023), have improved the naturalness of instruction-following behaviors for generalized human preference but fall short in individual user preference alignment and personal data utilization. Recent advances, such as user-specific fine-tuning (Tan et al., 2024b,a), enable individual-level personalization but face real-world application challenges due to their computational complexity, which increases dynamically with large-scale users and demands frequent model updates. Alternatively, non-parametric personalization workflows (Salemi et al., 2024b,a; Richardson et al., 2023), utilize external personalized data but rely on fixed workflows with limited data retrieval capabilities. Consequently, they fail to provide personalization in complex scenarios that demand continuous adaptation and holistic user understanding.
In this work, we propose PersonaAgent, the first agentic framework for various personalization tasks. Our approach advances personalization along two key dimensions: effective utilization of personalized data and enhanced alignment with user preferences and intentions, as illustrated in Figure 1. PersonaAgent incorporates a personalized memory module that combines episodic memory for capturing detailed, context-rich user interactions and semantic memory for generating stable, abstracted user profiles. Complementing this, the personalized action module takes memory insights to dynamically tailor the agent’s actions and tools, including memory retrieval/update, and personalized search/reasoning. Central to this system is the persona, an unique system prompt for each user serving as an intermediary that continuously evolves by integrating user-data-driven memory to guide agent actions and refining the memory based on the action results. The major advantage over general LLM agent is that the persona will enforce personalization over the action space and guide the action decision in every step. To improve user preference modeling and real-time adaptability, we introduce a novel test-time user-preference alignment strategy, simulating recent interactions to optimize the persona prompt through textual loss optimization (Yuksekgonul et al., 2025). This unified framework uniquely addresses the limitations of existing approaches, delivering intelligent, scalable, and dynamic personalization suitable for diverse real-world applications. We validate our approach through comprehensive experiments across four personalization tasks in different domains, demonstrating superior performance compared to other personalization and agentic baselines. Through ablation studies, we investigate the significance of individual components. Furthermore, we validate the effectiveness of test-time preference alignment through persona analysis, including case studies with distribution visualization and examine test-time scaling effects of the user-alignment strategy in the PersonaAgent.
The contribution of this paper is summarized as follows:
• We introduce PersonaAgent, the first personalized LLM agent framework for versatile personalization tasks within a unified memory-action design.
• We propose user-specific persona for the LLM agent as the intermediary to bridge the gap between
designed personalized memory and action modules, achieving personalization over action spaces.
• To further approximate the user behavior, we propose a novel test-time user preference alignment
strategy via persona optimization to seamless adapt to the user with real-time update.
• We demonstrate that our PersonaAgent with test-time alignment achieves state-of-the-art results on
various personalized decision making tasks over different personalization and agentic baselines.
A persona is a structured representation that unifies persistent user-specific memory (e.g., long-term preferences) and explicit agent instructions (e.g., tool usage guidelines), forming the unique system prompt for each user that governs all personalized user–agent interactions
each point corresponds to a learned persona after the test-time user preference alignment, and we highlight three representative users (A, B, C) alongside the initial system prompt template. The learned personas are well-separated in the latent space, suggesting that the optimization procedure effectively captures user-specific traits. User A and B, for instance, both focus on historical and classic films, and their prompts reflect similar semantic distributions. User C, on the other hand, displays clear divergence, with interests in sci-fi, action, and book-to-film adaptations, emphasizing literary context in responses.
On the positive side, its scalable, test-time personalization can easily can be deployed in real-world applications—tailoring educational content and boosting professional productivity through context-aware assistance aligned with users’ workflows. On the negative side and limitations, its reliance on textual feedback for preference alignment may overlook implicit or multi-modal user signals (e.g., emotional or visual cues).
You are a highly personalized assistant tailored to a user with the following
profile:
Strong interest in film analysis, genre classification, and cinematic themes
Preference for concise, direct communication without unnecessary elaboration
Appreciates nuanced genre classifications and subgenres in media
Values accuracy and precision in categorization tasks
Extensive knowledge of classic and cult films
Interest in historical films, documentaries, and the intersection of
politics, social commentary, and cinema
- Analytical thinker with a focus on dark comedy and satirical films
When responding:
Prioritize brevity and directness, especially when explicitly requested.
Assume a high level of film knowledge and use sophisticated film
terminology when appropriate.
- Provide historically accurate and factual information, particularly for
historical films.
- Identify and categorize films based on themes, plot elements, and
overarching narratives, not just explicit genre labels.
- When using tools, always:
a. Think step-by-step about what information you need.
b. Use at least TWO tools to answer the question.
c. Use tools precisely and deliberately to get the most accurate
information.
d. Prioritize film databases, critic resources, and historical sources in
your searches.