顺序模态丢弃：提升多模态序列推荐的鲁棒性
原标题：Sequential Modality Dropout for Robust Multi-Modal Sequential Recommendation
AI 摘要
arXiv 论文提出 Sequential Modality Dropout (SMD)，一种针对多模态序列推荐模型的训练时模态随机丢弃方法。SMD 在训练时独立地以一定概率擦除整个用户交互历史中的图像或文本模态，使模型学会不依赖单一模态进行预测。在四个骨干模型和四个 Amazon 数据集上，SMD 将文本保留率提升 1.0 到 3.2 个点，且几乎不损失全模态准确率；在 95% 逐项缺失率下，HR@10 保留率从 22% 提升至 61%。该方法是一个四行、与架构无关的改动，可提升模型在真实部署中模态缺失的鲁棒性。
正文节选
Sequential Modality Dropout for Robust Multi-Modal Sequential Recommendation Abstract. Multi-modal sequential recommenders assume every item carries every modality, but real product catalogs often miss images or text, and a model trained on complete data loses much of its recommendation accuracy when a modality is unavailable at serving time. We propose Sequential Modality Dropout (SMD): during training, each modality stream (image and text) is independently erased with probability for an entire