Generative Archetype-Grounded Item Representations for Sequential Recommendation

2026-06-09

AI Summary

Existing sequential recommendation systems predominantly rely on static semantic representations of items, which struggle to capture users’ dynamic behavioral patterns and overlook the role of target audiences in defining item identity. To address these limitations, this work proposes GenAIR, a novel framework that introduces generative audience personas for the first time. Specifically, GenAIR leverages large language models to generate semantic descriptions of an item’s ideal audience from its metadata and aligns these persona embeddings with real user interactions through a behavior calibration mechanism, thereby constructing behavior-aware item representations. This approach effectively bridges the gap between semantic space and user behavior, consistently outperforming state-of-the-art baselines and significantly enhancing the performance of various sequential recommendation models across three real-world datasets.

Abstract

Sequential recommendation aims to predict users' next interaction with items by analyzing their historical behavior. However, the limited quality of item representations remains a critical bottleneck. While pre-trained large language models (LLMs) can provide rich semantic representations, existing approaches only rely on static encoding of fixed attributes, overlooking the crucial role of target audiences in defining item identity. Moreover, the semantic space struggles to reflect actual user behavior, resulting in a significant gap between semantic representations and behavioral patterns. To address these limitations, we propose GenAIR, a general framework that empowers sequential recommendation with Generative Archetype-grounded Item Representations. Specifically, we first leverage an LLM to analyze item metadata and infer textual description of the Archetype, which represents the conceptual profile of the item's ideal target audience. We then extract the corresponding embeddings in a single forward pass. Further, to ground these generative archetypes in real-world behavior, we introduce a behavioral calibration objective, which explicitly incorporates behavioral signals from actual interactions. This objective adjusts the structure of the embedding space to reflect empirical patterns. GenAIR enables seamless integration with most existing models while maintaining high efficiency. Comprehensive experiments conducted on three real-world datasets demonstrate that GenAIR significantly improves the performance of various sequential recommendation models and consistently outperforms state-of-the-art baseline approaches. Implementation codes are available at https://github.com/AI-Santiago/GenAIR.

Problem

• sequential recommendation
• item representation
• semantic-behavior gap
• target audience
• large language models

Innovation

• Generative Archetype
• Item Representation
• Sequential Recommendation
• Behavioral Calibration
• Large Language Models