Developed and integrated FALCON-Bench, a benchmarking framework for evaluating multimodal large language models on one-hour video tasks, within the EvolvingLMMs-Lab/lmms-eval repository. The work focused on creating reproducible and scalable evaluation pipelines using Python scripting, with YAML-driven configurations to streamline task processing and model assessment. Utility functions were implemented to automate and standardize the evaluation process, enabling fair cross-model comparisons and supporting rapid iteration for research teams. The approach emphasized robust data processing and benchmarking practices, resulting in a maintainable framework that enhances the reliability of multimodal LLM evaluation without introducing major bug fixes during the development period.