融合多模态嵌入与序列建模，显著提升推荐系统点击率预测性能
1$^{st}$ Place Solution of WWW 2025 EReL@MIR Workshop Multimodal CTR Prediction Challenge
- 将冻结的多模态嵌入拼接至物品嵌入，简单高效整合多源信息
- 在挑战数据集上达成0.9839的AUC，远超基线模型表现
- 适合关注多模态推荐与在线点击率优化的研究者与工程师
WWW 2025 EReL@MIR Workshop 多模态点击率预测挑战赛聚焦于如何有效利用多模态嵌入特征以提升推荐系统的点击率预测能力。本文介绍我们在任务2中的第一名解决方案，结合序列建模与特征交互学习，充分捕捉用户-物品交互关系。对于多模态信息融合，我们仅将冻结的多模态嵌入拼接到每个物品嵌入后。在挑战数据集上的实验表明，该方法具有显著有效性，最终在排行榜上取得0.9839的AUC，远超基线模型。代码与配置已在GitHub公开，模型检查点可在HuggingFace获取。
原文摘要 · Abstract (English)
The WWW 2025 EReL@MIR Workshop Multimodal CTR Prediction Challenge focuses on effectively applying multimodal embedding features to improve click-through rate (CTR) prediction in recommender systems. This technical report presents our 1$^{st}$ place winning solution for Task 2, combining sequential modeling and feature interaction learning to effectively capture user-item interactions. For multimodal information integration, we simply append the frozen multimodal embeddings to each item embedding. Experiments on the challenge dataset demonstrate the effectiveness of our method, achieving superior performance with a 0.9839 AUC on the leaderboard, much higher than the baseline model. Code and configuration are available in our GitHub repository and the checkpoint of our model can be found in HuggingFace.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。