StePO-Rec: Towards Personalized Outfit Styling Assistant via Knowledge-Guided Multi-Step Reasoning

2025-04-14

AI Summary

Traditional fashion recommendation systems suffer from low transparency and insufficient integration of expert domain knowledge, limiting the effectiveness of personalized outfit recommendations. To address these challenges, this paper proposes PAFA—a multi-granularity fashion knowledge base—and StePO-Rec, a knowledge-guided, multi-step reasoning recommendation framework. PAFA establishes a three-level structured modeling hierarchy (scene–dimension–attribute), while StePO-Rec introduces a novel principle-aware knowledge organization scheme and a recursive tree-driven multi-step reasoning paradigm. Additionally, we design a style-consistency preservation mechanism that dynamically coordinates expert rules with user preferences. Our approach integrates semantic relation modeling and preference-trend-aware re-ranking. Evaluated on the IQON dataset, it achieves a 28% improvement in Recall@1 and a 32.8% gain in Mean Average Precision (MAP), significantly enhancing interpretability, recommendation reliability, and expert knowledge integration capability.

Technology Category

Knowledge Representation and Reasoning: PreferencesData Mining & Knowledge Management: Recommender SystemsMachine Learning: Learning Preferences or Rankings

Application Category

User Modeling, Personalization and Recommendation: Fairness-aware retrieval and rankingSearch and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingSemantics and Knowledge: Applications of semantic technologies for improving search, browsing, recommendation, personalization

Abstract

Advancements in Generative AI offers new opportunities for FashionAI, surpassing traditional recommendation systems that often lack transparency and struggle to integrate expert knowledge, leaving the potential for personalized fashion styling remain untapped. To address these challenges, we present PAFA (Principle-Aware Fashion), a multi-granular knowledge base that organizes professional styling expertise into three levels of metadata, domain principles, and semantic relationships. Using PAFA, we develop StePO-Rec, a knowledge-guided method for multi-step outfit recommendation. StePO-Rec provides structured suggestions using a scenario-dimension-attribute framework, employing recursive tree construction to align recommendations with both professional principles and individual preferences. A preference-trend re-ranking system further adapts to fashion trends while maintaining the consistency of the user's original style. Experiments on the widely used personalized outfit dataset IQON show a 28% increase in Recall@1 and 32.8% in MAP. Furthermore, case studies highlight improved explainability, traceability, result reliability, and the seamless integration of expertise and personalization.

Problem

• Enhance outfit recommendation transparency and expertise integration
• Develop multi-step personalized styling with knowledge guidance
• Improve fashion recommendation accuracy and explainability

Innovation

• Multi-granular knowledge base for styling expertise
• Recursive tree construction for outfit recommendations
• Preference-trend re-ranking system adaptation