ProMax: Exploring the Potential of LLM-derived Profiles with Distribution Shaping for Recommender Systems
Yi Zhang, Yiwen Zhang, Kai Zheng, Tong Chen, Hongzhi Yin
Abstract
The remarkable text understanding and generation capabilities of large language models (LLMs) have revitalized the field of general recommendation based on implicit user feedback. Rather than deploying LLMs directly as recommendation models, a more flexible paradigm leverages their ability to interpret users' historical interactions and semantic contexts to extract structured profiles that characterize user preferences. These profiles can be further transformed into actionable high-dimensional representations, serving as powerful signals to augment and strengthen recommendation models. However, the mechanism by which such profiles enhance recommendation performance within the feature space remains insufficiently understood. Moreover, existing studies predominantly rely on nonlinear alignment and fusion strategies to incorporate these profiles, which often lead to semantic loss and fail to fully exploit their potential. To address these limitations, we revisit profiles from a retrieval perspective and propose a simple yet effective recommendation framework built upon distribution shaping (ProMax) in this paper. We begin by employing dense retrieval to uncover the collaborative relationships between user and item profiles within the feature space. Based on this insight, we introduce a dual distribution-reshaping process, in which the profile distribution acts as a guiding signal to steer the recommendation model toward learning user preferences for unseen items beyond the scope of observed interactions. We apply ProMax to four classic recommendation methods on three public datasets. The results indicate that ProMax substantially improves base model performance and outperforms existing LLM-based recommendation approaches.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
- LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation cites Revisiting Graph Based Collaborative Filtering: A Linear Residual Graph Convolutional Network Approach
- Are Graph Augmentations Necessary?: Simple Graph Contrastive Learning for Recommendation cites LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation
Builds on24
- Learning Transferable Visual Models From Natural Language SupervisionAlec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh et al.ICML 2021 · 47,906 citations
- LoRA: Low-Rank Adaptation of Large Language ModelsEdward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu et al.ICLR 2022 · 18,833 citations
- LightGCN: Simplifying and Powering Graph Convolution Network for RecommendationXiangnan He, Kuan Deng, Xiang Wang, Yan Li et al.SIGIR 2020 · 4,448 citations
- Are Graph Augmentations Necessary?: Simple Graph Contrastive Learning for RecommendationJunliang Yu, Hongzhi Yin, Xin Xia, Tong Chen et al.SIGIR 2022 · 658 citations
- Revisiting Graph Based Collaborative Filtering: A Linear Residual Graph Convolutional Network ApproachLei Chen, Le Wu, Richang Hong, Kun Zhang et al.AAAI 2020 · 634 citations
Related papers
- ProEx: A Unified Framework Leveraging Large Language Model with Profile Extrapolation for RecommendationYi Zhang, Yiwen Zhang, Yu Wang, Tong Chen et al.KDD 2026 · 1 citation
- LettinGo: Explore User Profile Generation for Recommendation SystemLu Wang, Di Zhang, Fangkai Yang, Pu Zhao et al.KDD 2025 · 2 citations
- Representation Learning with Large Language Models for RecommendationXubin Ren, Wei Wei, Lianghao Xia, Lixin Su et al.WWW 2024 · 385 citations
- Rethinking LLM-Based Recommendations: A Personalized Query-Driven Parallel IntegrationDonghee Han, Hwanjun Song, Mun Yong YiEMNLP Findings 2025 · 1 citation
- Enhancing Large Language Model Based Sequential Recommender Systems with Pseudo Labels ReconstructionHyunsoo Na, Minseok Gang, Youngrok Ko, Jinseok Seol et al.EMNLP Findings 2024 · 3 citations
