The field of music generation and understanding is moving towards greater controllability and interpretability. Recent developments have focused on enabling fine-grained control over music generation, while also improving the understanding of musical concepts and attributes. This has been achieved through the use of novel frameworks and techniques, such as recursive feature machines and difficulty-aware curriculum learning. These advancements have shown promising results in tasks such as music generation, semantic understanding, and lyric translation. Notably, the ability to control and manipulate musical attributes, such as mood and genre, has been a key area of research. Overall, the field is advancing towards more sophisticated and controllable music generation and understanding systems. Noteworthy papers include: MuseTok, which proposes a tokenization method for symbolic music and achieves high-fidelity music reconstruction and accurate understanding of music theory. Steering Autoregressive Music Generation with Recursive Feature Machines, which introduces a framework for fine-grained control over pre-trained music models. LyriCAR, which proposes a difficulty-aware curriculum reinforcement learning framework for controllable lyric translation. Controllable Embedding Transformation for Mood-Guided Music Retrieval, which addresses the problem of controllable music retrieval through embedding-based transformation.