GRAM: Generative Recommendation via Semantic-aware Multi-granular Late Fusion

2025-06-02

AI Summary

Generative recommendation faces two key challenges: difficulty in modeling implicit item relationships and inefficient utilization of long textual information. To address these, we propose a semantic-aware multi-granularity post-fusion framework. Our approach introduces a novel “semantic → vocabulary” hierarchical relationship encoding mechanism to explicitly capture implicit inter-item associations. Additionally, we design a decoding-stage-delayed multi-granularity prompt fusion strategy, enabling hierarchical collaborative relationship modeling and late-fusion integration of heterogeneous prompts. Built upon the large language model (LLM)-based text generation paradigm, our method balances semantic richness with information completeness. Extensive experiments on four benchmark datasets demonstrate substantial improvements over eight state-of-the-art baselines: Recall@5 increases by 11.5–16.0%, and NDCG@5 improves by 5.3–13.6%.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Recommender Systems

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Generative recommendation is an emerging paradigm that leverages the extensive knowledge of large language models by formulating recommendations into a text-to-text generation task. However, existing studies face two key limitations in (i) incorporating implicit item relationships and (ii) utilizing rich yet lengthy item information. To address these challenges, we propose a Generative Recommender via semantic-Aware Multi-granular late fusion (GRAM), introducing two synergistic innovations. First, we design semantic-to-lexical translation to encode implicit hierarchical and collaborative item relationships into the vocabulary space of LLMs. Second, we present multi-granular late fusion to integrate rich semantics efficiently with minimal information loss. It employs separate encoders for multi-granular prompts, delaying the fusion until the decoding stage. Experiments on four benchmark datasets show that GRAM outperforms eight state-of-the-art generative recommendation models, achieving significant improvements of 11.5-16.0% in Recall@5 and 5.3-13.6% in NDCG@5. The source code is available at https://github.com/skleee/GRAM.

Problem

• Incorporating implicit item relationships into generative recommendation models
• Efficiently utilizing rich yet lengthy item information
• Improving generative recommendation performance via semantic-aware fusion

Innovation

• Semantic-to-lexical translation for item relationships
• Multi-granular late fusion for rich semantics
• Separate encoders for multi-granular prompts