Negative sampling plays a crucial role for cross-domain recommendation as it provides contrastive signals to learn user preference. Existing methods usually select items with high predicted scores or popularity as hard negative samples to improve model training. However, such methods suffer from choosing false negative samples since items with high predicted scores or popularity could also indicate potential positive user preference. Although several studies devoted to discovering true negative samples, few of them leverage user cross-domain behaviors to alleviate the false negative issue. How to effectively mine and utilize hard negative samples to improve cross-domain recommendation remains an open question.

Generative models have emerged as a promising utility to enhance recommender systems. It is essential to model both item content and user-item collaborative interactions in a unified generative framework for better recommendation. Although some existing large language model (LLM)-based methods contribute to fusing content information and collaborative signals, they fundamentally rely on textual language generation, which is not fully aligned with the recommendation task. How to integrate content knowledge and collaborative interaction signals in a generative framework tailored for item recommendation is still an open research challenge.

User profiling is one of the most important components in recommendation systems, where a user is profiled using demographic (e.g. gender, age, and location) and user behavior information (e.g. browsing and search history). Among different dimensions of user profiling, tagging is an explainable and widely-used representation of user interest. In this paper, we propose a user tag profiling model (UTPM) to study user-tag profiling as a multi-label classification task using deep neural networks. Different from the conventional model, our UTPM model is a multi-head attention mechanism with shared query vectors to learn sparse features across different fields. Besides, we introduce the improved FM-based cross feature layer, which outperforms many state-of-the-art cross feature methods and further enhances model performance. Meanwhile, we design a novel joint method to learn the preference of different tags from a single clicked news article in recommendation systems. Furthermore, our UTPM model is deployed in the WeChat "Top Stories" recommender system, where both online and offline experiments demonstrate the superiority of the proposed model over baseline models.