The field of recommendation systems is witnessing significant developments, with a focus on enhancing the accuracy and interpretability of recommendations. Researchers are exploring innovative approaches to disentangle user intentions, model complex user-item interactions, and leverage structural signals from multiple perspectives. Notably, generative models are being applied to recommendation tasks, enabling the development of foundation models that can excel across diverse tasks. Furthermore, collaborative filtering methods are being improved to capture fine-grained user-item compatibility and adapt to long-tail scenarios. The importance of modeling dual-level user interests, including both group-level and item-level interests, is also being recognized. Interestingly, some studies are investigating the application of lookalike algorithms and next-user modeling to address cold-start challenges. Additionally, surveys are being conducted to systematically review the progress and challenges in multi-interest recommendation. Some noteworthy papers include: Dual-View Disentangled Multi-Intent Learning for Enhanced Collaborative Filtering, which proposes a unified framework for disentangling user intentions and modeling interaction-level intent alignment. Generative Representational Learning of Foundation Models for Recommendation, which introduces a novel framework for training recommendation foundation models. DiscRec: Disentangled Semantic-Collaborative Modeling for Generative Recommendation, which proposes a novel framework for disentangled semantic-collaborative signal modeling.