返回全部动态
面向LLM序列推荐的鲁棒个性化提示学习
原标题：Learning Robust Personalized Prompts for LLM-Driven Sequential Recommendation
AI 摘要
该论文提出 LRPRec，一个面向 LLM 驱动序列推荐的可学习提示框架。针对现有方法对提示措辞高度敏感、缺乏个性化以及连续提示易发生语义漂移的问题，LRPRec 将指令提示参数化为由离散模板初始化的连续向量，并引入个性化提示注入机制与语义漂移约束。在三个基准数据集上，该方法在无需人工调优背景与任务推理模板的情况下，稳定优于强基线。
以上摘要由 AI 生成，可能存在误差。事实请以原文为准。
正文节选
Learning Robust Personalized Prompts for LLM-Driven Sequential Recommendation Abstract. With the rapid advancement of large language models (LLMs), LLM-driven sequential recommendation has emerged as a promising paradigm that formulates next-item prediction as autoregressive generation conditioned on natural-language prompts. Despite their potential, existing methods exhibit high sensitivity to prompt formulations: even semantically equivalent prompts with minor wording variations can lead to su
发布时间：2026-10-06 12:00
抓取时间：2026-10-06 12:52
来源机构：arXiv