Prospect Personalized Recommendation on Large Language Model-based Agent Platform

TL;DR

The paper addresses embedding recommender systems within Large Language Model–based Agent platforms, where agents are interactive, intelligent, and proactive. It introduces Rec4Agentverse, a paradigm built on two roles—Agent Item and Agent Recommender—and outlines a three-stage evolution to progressively enhance information flow among users, items, and recommenders. A preliminary travel-planning demonstration illustrates feasibility and the potential for richer, personalized services through cross-agent collaboration. The work identifies key research directions, risks, and practical considerations, highlighting the need for quantitative evaluation and broader deployment to advance AI-enabled information systems.

Abstract

The new kind of Agent-oriented information system, exemplified by GPTs, urges us to inspect the information system infrastructure to support Agent-level information processing and to adapt to the characteristics of Large Language Model (LLM)-based Agents, such as interactivity. In this work, we envisage the prospect of the recommender system on LLM-based Agent platforms and introduce a novel recommendation paradigm called Rec4Agentverse, comprised of Agent Items and Agent Recommender. Rec4Agentverse emphasizes the collaboration between Agent Items and Agent Recommender, thereby promoting personalized information services and enhancing the exchange of information beyond the traditional user-recommender feedback loop. Additionally, we prospect the evolution of Rec4Agentverse and conceptualize it into three stages based on the enhancement of the interaction and information exchange among Agent Items, Agent Recommender, and the user. A preliminary study involving several cases of Rec4Agentverse validates its significant potential for application. Lastly, we discuss potential issues and promising directions for future research.

Figures (7)

• Figure 1: An example of interaction between an Agent Item and a user. A Travel Agent can serve as an information carrier with travel-related information, as well as engage in a dialogue with the user to exchange related information.
• Figure 2: Illustration of the Rec4Agentverse paradigm. The left portion of the diagram depicts three roles in RecAgentverse: user, Agent Recommender, and Agent Item, along with their interconnected relationships. In contrast to traditional recommender systems, Rec4Agentverse has more intimate relationships among the three roles. For instance, there are multi-round interactions between 1) users and Agent Items, and 2) Agent Recommender and Agent Items. The right side of the diagram demonstrates that Agent Recommender can collaborate with Agent Items to affect the information flow of users and offer personalized information services.
• Figure 3: Three stages of Rec4Agentverse. The bidirectional arrows depicted in the Figure symbolize the flow of information. During the first stage of User-Agent interaction, information flows between the user and Agent Item. In the Agent-Recommender Collaboration stage, information flows between Agent Item and Agent Recommender. For the Agents Collaboration stage, information flows between various Agent Items.
• Figure 4: The three stages of our proposed Rec4Agentverse paradigm. (a) For the User-Agent interaction stage, users can interact efficiently with Agent items through natural language. (b) For the Agent-Recommender collaboration stage, Agent Item and Agent Recommender could interact with each other. "Evolvement" means that the preference of the user can also be used for Agent Item to evolve itself or to evolve itself with the help of Agent Recommender. "Agent Feedback" refers to that the recommended Agent Item can feed the preference of the user back to Agent Recommender. "Proactive" stands for Agent Recommender can send information or issue instructions to Agent items. (c) For the Agents collaboration stage, Agent Items can collaborate together to provide personalized information services for the user.
• Figure 5: A case of the User-Agent interaction stage. The user expressed the desire for the Travel Agent to Agent Recommender and get back a recommendation. Subsequently, the user and the Travel Agent engaged in interactions to make the travel plan.