Personalization Matters: Interactive Conversational Shopping Agent with Dialogue User-Centric RAG

ACL ARR 2026 January Submission3763 Authors

Keywords: Personlized Agents; Dialogue Interaction; LLM Interaction; Shopping Agents

Abstract: Real-world shopping requires dynamic interactions that surpass static search capabilities. Large Language Models (LLMs) function as "Personalization Agents" to align responses with individual preferences; however, effective frameworks and realistic benchmarks for domain-specific shopping remain scarce. This paper proposes a novel multi-agent framework addressing personalized shopping requirements, including user preferences and past interactions. We propose a novel multi-agent framework utilizing multimodal Retrieval-Augmented Generation (RAG) to integrate product metadata and user profiles for personalized shopping. Addressing the lack of domain-specific benchmarks, we introduce a robust evaluation suite for intent alignment and dialogue consistency. Experiments demonstrate significant improvements in preference tracking, cumulative information synthesis, and interaction progression compared to baselines.

Paper Type: Long

Research Area: AI/LLM Agents

Research Area Keywords: interactive agents, dialogue systems for shopping domain, evaluation and metrics, interactive conversation, multi-modal agents, RAG

Contribution Types: NLP engineering experiment, Reproduction study, Approaches to low-resource settings, Approaches low compute settings-efficiency, Data resources, Data analysis

Languages Studied: English

Submission Number: 3763