The field of multimodal sentiment analysis and emotion recognition is moving towards addressing the challenges of missing or inconsistent modalities. Researchers are proposing innovative frameworks and models that can effectively handle these issues, such as factorization-guided semantic recovery, tri-modal severity fused diagnosis, and calibrated multimodal consensus. These approaches aim to improve the accuracy and robustness of sentiment analysis and emotion recognition systems. Noteworthy papers in this area include: FSRF, which proposes a factorization-guided semantic recovery framework to mitigate the modality missing problem. Tri-Modal Severity Fused Diagnosis, which presents a unified tri-modal affective severity framework for diagnosing depression and PTSD. Calibrating Multimodal Consensus for Emotion Recognition, which introduces a model that addresses semantic inconsistencies across modalities. SheafAlign, which proposes a sheaf-theoretic framework for decentralized multimodal alignment.