Developed and integrated the GroundCocoa benchmark task into both the red-hat-data-services/lm-evaluation-harness and swiss-ai/lm-evaluation-harness repositories, expanding evaluation capabilities for compositional and conditional reasoning in flight booking scenarios. The work involved designing new YAML-based task configurations, implementing Python processing utilities, and updating Markdown documentation to support adoption and contributor onboarding. By focusing on benchmark development and data processing, the contributions enhanced model assessment workflows and prepared the codebase for future scaling. The updates improved documentation clarity and ensured that the evaluation harnesses could accommodate domain-specific reasoning tasks, reflecting a methodical approach to machine learning evaluation and natural language processing.