Where to Go Next for Recommender Systems? ID- vs. Modality-based Recommender Models Revisited

By Fajie Yuan, Yu Song, Youhua Li, Junchen Fu, Fei Yang, Yunzhu Pan, Yongxin Ni. Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, 2023

341 citations recorded by the source corpus.

Recommendation models that utilize unique identities (IDs for short) to represent distinct users and items have been state-of-the-art (SOTA) and dominated the recommender systems (RS) literature for over a decade. Meanwhile, the pre-trained modality encoders, such as BERT [9] and Vision Transformer [11], have become increasingly powerful in modeling the raw modality features of an item, such as text and images. Given this, a natural question arises: can a purely modality-based recommendation model (MoRec) outperforms or matches a pure ID-based model (IDRec) by replacing the itemID embedding with a SOTA modality encoder? In fact, this question was answered ten years ago when IDRec beats MoRec by a strong margin in both recommendation accuracy and efficiency. We aim to revisit this 'old' question and systematically study MoRec from several aspects. Specifically, we study several sub-questions: (i) which recommendation paradigm, MoRec or IDRec, performs better in practical scenarios, especially in the general setting and warm item scenarios where IDRec has a strong advantage? does this hold for items with different modality features? (ii) can the latest technical advances from other communities (i.e., natural language processing and computer vision) translate into accuracy improvement for MoRec? (iii) how to effectively utilize item modality representation, can we use it directly or do we have to adjust it with new data? (iv) are there any key challenges that MoRec needs to address in practical applications? To answer them, we conduct rigorous experiments for item recommendations with two popular modalities, i.e., text and vision. We provide the first empirical evidence that MoRec is already comparable to its IDRec counterpart with an expensive end-to-end training method, even for warm item recommendation. Our results potentially imply that the dominance of IDRec in the RS field may be greatly challenged in the future. We release our code and other materials at https://github.com/westlake-repl/IDvs.MoRec.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Information Systems

View Paper: https://www.semanticscholar.org/paper/11628f656257e75e46447ac21cdaa86c4b340a0a

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Xiangnan He
2. Yi Tay
3. Fuli Feng
4. Yupeng Hou
5. Julian McAuley
6. Chao Huang
7. Jiacheng Li
8. Lianghao Xia

In Computer Vision and Pattern Recognition

1. Xiangnan He
2. Fuli Feng
3. Yi Tay
4. Yupeng Hou
5. Julian McAuley
6. Jiacheng Li
7. Chao Huang
8. Wenjie Wang

In Information Systems

1. Xiangnan He
2. Fuli Feng
3. Yupeng Hou
4. Yi Tay
5. Chao Huang
6. Jiacheng Li
7. Lianghao Xia
8. Julian McAuley

Citations in the corpus

• Bridging Language and Items for Retrieval and Recommendation: Benchmarking LLMs as Semantic Encoders (2024, positive): “Consequently, an increasing number of studies has explored using LLM-encoded text representations as inputs to train recommendation models, a direction referred to as modality-based recommendation (Hou et al., 2022; Yuan et al., 2023).”
• A survey on large language models for recommendation (2023, neutral): “, Yuan et al. (2023) conducted large-scale experiments and found that modern MoRec could already perform on par or better than IDRec with the typical recommendation architecture (i.e., Transformer backbone) even in the non-cold-start item recommendation setting with the SOTA and E2E-trained…”
• Recommender Systems with Generative Retrieval (2023, methodology): “Figure 2: A high-level overview of the modeling approach used in TIGER.”
• When large language models meet personalization: perspectives of challenges and opportunities (2023, neutral): “Additionally, some recent works [114–117] have explored using only textual features as inputs to recommendation models, leveraging pre-trained language models to alleviate cold-start problems and enable cross-domain recommendations.”
• TALLRec: An Effective and Efficient Tuning Framework to Align Large Language Model with Recommendation (2023, background): “With the advancement of LLMs, people are gradually recognizing their potential in recommendation systems [30, 63, 64].”
• Recommender Systems in the Era of Large Language Models (LLMs) (2023, neutral): “Details of the two recommender systems are illustrated in Figure 2.”
• LLMRec: Large Language Models with Graph Augmentation for Recommendation (2023, background): “Firstly , LLMRec embraces the shift from an ID-based recommendation framework to a modality-based paradigm [18, 46].”
• Representation Learning with Large Language Models for Recommendation (2023, background): “In recent times, there have been several endeavors to leverage diverse data modalities in order to enhance traditional ID-based recommenders [9, 17, 46].”