arXiv：cs.CL· Xinfeng Wang, Jin Cui, Fumiyo Fukumoto, Yoshimi Suzuki·· 10 小时前AI 评分29
ELMRec：增强 LLM 推荐模型的高阶交互感知能力
Enhancing High-order Interaction Awareness in LLM-based Recommender Model
AI 导读
针对现有 LLM 推荐方法忽视或无法有效建模用户-物品高阶交互的问题，研究者提出 ELMRec，通过增强 whole-word embedding 让 LLM 理解图结构交互，且无需图预训练。研究还发现 LLM 倾向依据用户早期交互而非近期交互推荐物品，并提出重排序方案。ELMRec 在直接推荐和序列推荐上均优于 SOTA 方法，论文已被 EMNLP 2024 Main 接收。
正文
Abstract:Large language models (LLMs) have demonstrated prominent reasoning capabilities in recommendation tasks by transforming them into text-generation tasks. However, existing approaches either disregard or ineffectively model the user-item high-order interactions. To this end, this paper presents an enhanced LLM-based recommender (ELMRec). We enhance whole-word embeddings to substantially enhance LLMs' interpretation of graph-constructed interactions for recommendations, without requiring graph pre-training. This finding may inspire endeavors to incorporate rich knowledge graphs into LLM-based recommenders via whole-word embedding. We also found that LLMs often recommend items based on users' earlier interactions rather than recent ones, and present a reranking solution. Our ELMRec outperforms state-of-the-art (SOTA) methods in both direct and sequential recommendations.
| Comments: | Long paper accepted to EMNLP 2024 Main. 16 pages |
| Subjects: | Information Retrieval (cs.IR); Computation and Language (cs.CL) |
| Cite as: | arXiv:2409.19979 [cs.IR] |
| (or arXiv:2409.19979v4 [cs.IR] for this version) | |
| https://doi.org/10.48550/arXiv.2409.19979 arXiv-issued DOI via DataCite |
Submission history
From: Xinfeng Wang [view email]
[v1]
Mon, 30 Sep 2024 06:07:12 UTC (6,767 KB)
[v2]
Tue, 1 Oct 2024 13:04:55 UTC (6,767 KB)
[v3]
Mon, 18 Nov 2024 06:28:01 UTC (6,768 KB)
[v4]
Tue, 6 Oct 2026 12:59:24 UTC (3,814 KB)
来源：arXiv：cs.CL · arxiv.org