Lost in Sequence: Do Large Language Models Understand Sequential Recommendation?
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910999627431936 |
|---|---|
| author | Kim, Sein Kang, Hongseok Kim, Kibum Kim, Jiwan Kim, Donghyun Yang, Minchul Oh, Kwangjin McAuley, Julian Park, Chanyoung |
| author_facet | Kim, Sein Kang, Hongseok Kim, Kibum Kim, Jiwan Kim, Donghyun Yang, Minchul Oh, Kwangjin McAuley, Julian Park, Chanyoung |
| contents | Large Language Models (LLMs) have recently emerged as promising tools for recommendation thanks to their advanced textual understanding ability and context-awareness. Despite the current practice of training and evaluating LLM-based recommendation (LLM4Rec) models under a sequential recommendation scenario, we found that whether these models understand the sequential information inherent in users' item interaction sequences has been largely overlooked. In this paper, we first demonstrate through a series of experiments that existing LLM4Rec models do not fully capture sequential information both during training and inference. Then, we propose a simple yet effective LLM-based sequential recommender, called LLM-SRec, a method that enhances the integration of sequential information into LLMs by distilling the user representations extracted from a pre-trained CF-SRec model into LLMs. Our extensive experiments show that LLM-SRec enhances LLMs' ability to understand users' item interaction sequences, ultimately leading to improved recommendation performance. Furthermore, unlike existing LLM4Rec models that require fine-tuning of LLMs, LLM-SRec achieves state-of-the-art performance by training only a few lightweight MLPs, highlighting its practicality in real-world applications. Our code is available at https://github.com/Sein-Kim/LLM-SRec. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_13909 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Lost in Sequence: Do Large Language Models Understand Sequential Recommendation? Kim, Sein Kang, Hongseok Kim, Kibum Kim, Jiwan Kim, Donghyun Yang, Minchul Oh, Kwangjin McAuley, Julian Park, Chanyoung Information Retrieval Artificial Intelligence Large Language Models (LLMs) have recently emerged as promising tools for recommendation thanks to their advanced textual understanding ability and context-awareness. Despite the current practice of training and evaluating LLM-based recommendation (LLM4Rec) models under a sequential recommendation scenario, we found that whether these models understand the sequential information inherent in users' item interaction sequences has been largely overlooked. In this paper, we first demonstrate through a series of experiments that existing LLM4Rec models do not fully capture sequential information both during training and inference. Then, we propose a simple yet effective LLM-based sequential recommender, called LLM-SRec, a method that enhances the integration of sequential information into LLMs by distilling the user representations extracted from a pre-trained CF-SRec model into LLMs. Our extensive experiments show that LLM-SRec enhances LLMs' ability to understand users' item interaction sequences, ultimately leading to improved recommendation performance. Furthermore, unlike existing LLM4Rec models that require fine-tuning of LLMs, LLM-SRec achieves state-of-the-art performance by training only a few lightweight MLPs, highlighting its practicality in real-world applications. Our code is available at https://github.com/Sein-Kim/LLM-SRec. |
| title | Lost in Sequence: Do Large Language Models Understand Sequential Recommendation? |
| topic | Information Retrieval Artificial Intelligence |
| url | https://arxiv.org/abs/2502.13909 |