Meta 用生成式模型探索 Facebook Forum 帖子检索
原标题:Exploring Forum Post Retrieval with Generative Modeling
AI 摘要
Meta 在 arXiv 发表研究,探索用生成式推荐(GR)为 Facebook Groups 的新独立应用 Forum 做帖子检索。由于 Forum 自身交互数据稀疏,团队采用跨平台迁移:在更大的 Facebook Groups 互动语料上训练,并复用基于 Facebook Feed 数据训练的 RQ-VAE 分层前缀语义 ID(SID),再用 3B 参数指令微调语言模型直接从用户上下文生成 SID。研究系统消融了 SID 构造、用户历史组成与长度、用户画像特征等设计选择,发现 SID 深度影响最大,且最小的基座模型最具成本效益。
正文节选
Exploring Forum Post Retrieval with Generative Modeling Abstract Generative recommendation (GR) has emerged as an alternative to embedding-based retrieval, building on the success of generative models in language and vision. We are exploring GR on Facebook Forum, a standalone application for medium-to-heavy users of Facebook Groups. Because Forum is a new surface, its own interaction data are too sparse to train a GR model from scratch. We address this with transfer along two axes: we train on a