Large Language Models meet Collaborative Filtering: An Efficient All-round LLM-based Recommender System

TL;DR

A-LLMRec presents an all-round LLM-based recommender that directly leverages collaborative knowledge from a frozen CF-RecSys by learning an alignment network to map item embeddings into the LLM’s token space. The framework operates in two stages: Stage-1 aligns collaborative item embeddings with textualSBERT representations to form joint collaborative-text embeddings, using reconstruction and recommendation losses to prevent information loss; Stage-2 projects these joint embeddings and user representations into the LLM’s token space and designs prompts for the LLM to generate recommendations without fine-tuning the LLM. Empirical results on four Amazon datasets show A-LLMRec outperforms traditional CF, modality-aware, and prior LLM-based methods across cold/warm, few-shot, cold-user, and cross-domain scenarios, while offering model-agnostic integration and faster training/inference than LLM-fine-tuning baselines. In addition to recommendations, A-LLMRec demonstrates language-generation capabilities (favorite-genre predictions), underscoring the approach’s broader applicability and practical impact for scalable, real-world systems.

Abstract

Collaborative filtering recommender systems (CF-RecSys) have shown successive results in enhancing the user experience on social media and e-commerce platforms. However, as CF-RecSys struggles under cold scenarios with sparse user-item interactions, recent strategies have focused on leveraging modality information of user/items (e.g., text or images) based on pre-trained modality encoders and Large Language Models (LLMs). Despite their effectiveness under cold scenarios, we observe that they underperform simple traditional collaborative filtering models under warm scenarios due to the lack of collaborative knowledge. In this work, we propose an efficient All-round LLM-based Recommender system, called A-LLMRec, that excels not only in the cold scenario but also in the warm scenario. Our main idea is to enable an LLM to directly leverage the collaborative knowledge contained in a pre-trained state-of-the-art CF-RecSys so that the emergent ability of the LLM as well as the high-quality user/item embeddings that are already trained by the state-of-the-art CF-RecSys can be jointly exploited. This approach yields two advantages: (1) model-agnostic, allowing for integration with various existing CF-RecSys, and (2) efficiency, eliminating the extensive fine-tuning typically required for LLM-based recommenders. Our extensive experiments on various real-world datasets demonstrate the superiority of A-LLMRec in various scenarios, including cold/warm, few-shot, cold user, and cross-domain scenarios. Beyond the recommendation task, we also show the potential of A-LLMRec in generating natural language outputs based on the understanding of the collaborative knowledge by performing a favorite genre prediction task. Our code is available at https://github.com/ghdtjr/A-LLMRec .

Figures (6)

• Figure 1: Comparisons between collaborative filtering model (SASRec), modality-aware model (i.e., MoRec), and LLM-based model (i.e., TALLRec) under the cold/warm scenarios on Amazon Movies/Video Games dataset (Hit@1).
• Figure 2: (a) is the overview of A-LLMRec. (b) and (c) are the detailed architecture of Stage 1 and Stage 2, respectively.
• Figure 3: An example prompt of A-LLMRec designed for the Amazon Movies dataset. For other datasets, we keep the same format but adjust the verbs and nouns to fit the context (e.g., 'watched' $\rightarrow$ 'bought', 'movie' $\rightarrow$ 'item').
• Figure 4: A-LLMRec v.s. LLM-Only on the favorite genre prediction task (Movies and TV dataset used).
• Figure 5: A-LLMRec, LLM-Only, and TALLRec on the favorite genre prediction task (Movies and TV dataset used).