The field of video understanding is experiencing significant growth, with a focus on developing more efficient and effective methods for analyzing long videos. Researchers are exploring innovative approaches to improve the performance of vision-language models, including the use of uncertainty-aware chain-of-thought processes, relevance-aware adapters, and relation-based video representation learning. These advancements have led to notable improvements in video question answering, video classification, and video captioning tasks. Noteworthy papers include: VideoAgent2, which proposes an uncertainty-aware CoT process for long video analysis, achieving a 13.1% average improvement over the previous state-of-the-art method. REEF, which introduces an efficient LLM adapter for video-level understanding, reducing computational overhead by up to 34% while achieving competitive results. REVEAL, which presents a relation-based video representation learning framework, capturing visual relation information and achieving competitive results on five challenging benchmarks. LVC, which proposes a lightweight compression framework for enhancing VLMs in long video understanding, providing consistent performance improvements across various models.