Worked on integrating the MMVet-v2 task into the lmms-eval evaluation framework, focusing on enhancing image processing and configuration management for visual data workflows. Leveraged Python and YAML to update configuration files, introducing an interleaved_format flag and standardizing image token handling to improve the fidelity of visual data integration. The work included validating the end-to-end evaluation pipeline, ensuring that new configuration and image processing utilities functioned seamlessly within the repository. Emphasized stability and maintainability by delivering these features without introducing new bugs, demonstrating a methodical approach to evolving data processing and image handling capabilities in the lmms-eval project.