Lifelong Personalized Low-rank Adaptation Of Large Language Models For Recommendation

Zhu Jiachen, Lin Jianghao, Dai Xinyi, Chen Bo, Shan Rong, Zhu Jieming, Tang Ruiming, Yu Yong, Zhang Weinan. Arxiv 2024

[Paper]    
Efficiency And Optimization Fine Tuning Reinforcement Learning Training Techniques

We primarily focus on the field of large language models (LLMs) for recommendation, which has been actively explored recently and poses a significant challenge in effectively enhancing recommender systems with logical reasoning abilities and open-world knowledge. Current mainstream efforts mainly center around injecting personalized information from recommendation models into LLMs by customizing input templates or aligning representations between semantic and recommendation spaces at the prediction layer. However, they face three significant limitations: (1) LoRA is mostly used as a core component in existing works, but personalization is not well established in LoRA parameters as the LoRA matrix shared by every user may not cater to different users’ characteristics, leading to suboptimal performance. (2) Although lifelong personalized behavior sequences are ideal for personalization, their use raises effectiveness and efficiency issues since LLMs require escalating training and inference time to extend text lengths. (3) Existing approaches aren’t scalable for large datasets due to training efficiency constraints. Thus, LLMs only see a small fraction of the datasets (e.g., less than 10%) instead of the whole datasets, limiting their exposure to the full training space. To address these problems, we propose RecLoRA. This model incorporates a Personalized LoRA module that maintains independent LoRAs for different users and a Long-Short Modality Retriever that retrieves different history lengths for different modalities, significantly improving performance while adding minimal time cost. Furthermore, we design a Few2Many Learning Strategy, using a conventional recommendation model as a lens to magnify small training spaces to full spaces. Extensive experiments on public datasets demonstrate the efficacy of our RecLoRA compared to existing baseline models.

Similar Work