AgentDR Dynamic Recommendation with Implicit Item-Item Relations via LLM-based Agents
This work addresses the challenge of improving recommendation relevance and scalability for users in e-commerce by leveraging LLMs to reason over implicit item relationships, representing an incremental advancement over existing agent-based methods.
The paper tackles the problem of hallucination and scalability in agent-based recommendation systems by proposing AgentDR, a framework that combines LLM reasoning with traditional recommendation tools to capture implicit item-item relationships, achieving on average a twofold improvement in full-ranking performance across three grocery datasets.
Recent agent-based recommendation frameworks aim to simulate user behaviors by incorporating memory mechanisms and prompting strategies, but they struggle with hallucinating non-existent items and full-catalog ranking. Besides, a largely underexplored opportunity lies in leveraging LLMs'commonsense reasoning to capture user intent through substitute and complement relationships between items, which are usually implicit in datasets and difficult for traditional ID-based recommenders to capture. In this work, we propose a novel LLM-agent framework, AgenDR, which bridges LLM reasoning with scalable recommendation tools. Our approach delegates full-ranking tasks to traditional models while utilizing LLMs to (i) integrate multiple recommendation outputs based on personalized tool suitability and (ii) reason over substitute and complement relationships grounded in user history. This design mitigates hallucination, scales to large catalogs, and enhances recommendation relevance through relational reasoning. Through extensive experiments on three public grocery datasets, we show that our framework achieves superior full-ranking performance, yielding on average a twofold improvement over its underlying tools. We also introduce a new LLM-based evaluation metric that jointly measures semantic alignment and ranking correctness.