The field of vision-language models and large language models is rapidly advancing, with a focus on long-context understanding and generation. Recent developments have highlighted the importance of evaluating these models' ability to handle complex, long-range semantic dependencies and follow explicit length instructions. Researchers are introducing new benchmarks and evaluation frameworks to assess the performance of these models, such as MMLongBench for long-context vision-language tasks and LIFEBench for length instruction following. These efforts aim to provide a more comprehensive understanding of the strengths and limitations of current models and guide future development. Notably, some papers are pushing the boundaries of long-context understanding, such as Too Long, Didn't Model, which tests a model's ability to report plot summary and storyworld configuration in novels. Meanwhile, others like WebNovelBench are exploring the evaluation of long-form storytelling capabilities of large language models. Overall, the field is moving towards more nuanced and comprehensive evaluation of long-context vision-language models and large language models. Notable papers include: MMLongBench, which provides a comprehensive analysis of long-context vision-language models. WebNovelBench, which introduces a novel benchmark for evaluating long-form novel generation. Too Long, Didn't Model, which releases a benchmark testing a model's ability to report plot summary and storyworld configuration in novels. LIFEBench, which comprehensively evaluates large language models' ability to follow length instructions. CASTILLO, which characterizes response length distributions of large language models.