The field of multimodal video analysis is moving towards more comprehensive and fine-grained assessments of video understanding and retrieval. Researchers are developing new benchmarks and evaluation criteria to probe the capabilities of multimodal large language models (MLLMs) in tasks such as video fake news detection and video aesthetic assessment. These efforts aim to advance the state-of-the-art in video analysis by providing more nuanced and detailed evaluations of model performance. Notable papers in this area include: MUVR, which proposes a new benchmark for multi-modal untrimmed video retrieval, and A Video Is Not Worth a Thousand Words, which introduces a method for computing feature attributions and modality scores in multimodal models. Perception, Understanding and Reasoning, and VADB also present significant contributions to the field, with the former introducing a benchmark for video fake news detection and the latter presenting a large-scale video aesthetic database with professional annotations.