The field of affective computing and recommendation systems is moving towards a more nuanced understanding of human emotions and behaviors. Researchers are developing innovative frameworks and models that can disentangle complex emotional dynamics, capture dynamic user preferences, and mitigate the spread of misinformation. A key direction is the integration of hierarchical structural mechanisms, low-rank sparse modeling, and continuous-time discrete-space diffusion processes to improve the robustness and accuracy of affective analysis and recommendation systems. Another important trend is the development of closed-loop control frameworks that can balance user engagement with misinformation suppression and content moderation. Noteworthy papers include:

• Disentangling Emotional Bases and Transient Fluctuations, which proposes a Low-Rank Sparse Emotion Understanding Framework for video affective analysis.
• Continuous-time Discrete-space Diffusion Model for Recommendation, which introduces a novel diffusion-based recommendation framework that operates in discrete space and continuous time.
• Learning to Control Misinformation, which presents a control framework that mitigates misinformation spread while maintaining user engagement.
• DualGR, which proposes a generative retrieval framework that explicitly models dual horizons of user interests with selective activation.
• Causally-Informed Reinforcement Learning for Adaptive Emotion-Aware Social Media Recommendation, which integrates a Transformer-based emotion predictor with a hybrid recommendation policy to personalize content based on users' evolving emotional trajectories.
• Revisiting Fairness-aware Interactive Recommendation, which introduces a novel control knob, i.e., the lifecycle of items, to dynamically harmonize fairness and accuracy in interactive recommendation systems.