# Multi-Relational Contrastive Learning for Recommendation

arXiv: https://arxiv.org/abs/2309.01103 · HTML: https://mlsanity.com/paper/2309.01103/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 3 Sep 2023.

## Built on by (most cited first, 5 of 5)

- [A Comprehensive Survey on Self-Supervised Learning for Recommendation](https://mlsanity.com/paper/2404.03354/) (2024)
- [Dual-Channel Multiplex Graph Neural Networks for Recommendation](https://mlsanity.com/paper/2403.11624/) (2024)
- [Dual-Phase Playtime-guided Recommendation: Interest Intensity Exploration and Multimodal Random Walks](https://mlsanity.com/paper/2508.14058/) (2025)
- [Rejuvenating Cross-Entropy Loss in Knowledge Distillation for Recommender Systems](https://mlsanity.com/paper/2509.20989/) (2025)
- [SLSREC: Self-Supervised Contrastive Learning for Adaptive Fusion of Long- and Short-Term User Interests](https://mlsanity.com/paper/2604.04530/) (2026)

## Abstract (arXiv)

Personalized recommender systems play a crucial role in capturing users' evolving preferences over time to provide accurate and effective recommendations on various online platforms. However, many recommendation models rely on a single type of behavior learning, which limits their ability to represent the complex relationships between users and items in real-life scenarios. In such situations, users interact with items in multiple ways, including clicking, tagging as favorite, reviewing, and purchasing. To address this issue, we propose the Relation-aware Contrastive Learning (RCL) framework, which effectively models dynamic interaction heterogeneity. The RCL model incorporates a multi-relational graph encoder that captures short-term preference heterogeneity while preserving the dedicated relation semantics for different types of user-item interactions. Moreover, we design a dynamic cross-relational memory network that enables the RCL model to capture users' long-term multi-behavior preferences and the underlying evolving cross-type behavior dependencies over time. To obtain robust and informative user representations with both commonality and diversity across multi-behavior interactions, we introduce a multi-relational contrastive learning paradigm with heterogeneous short- and long-term interest modeling. Our extensive experimental studies on several real-world datasets demonstrate the superiority of the RCL recommender system over various state-of-the-art baselines in terms of recommendation accuracy and effectiveness.