Worked on the docling-project/docling-eval repository to enhance the accuracy and usability of benchmark documentation. Focused on correcting benchmark naming in FinTabNet documentation and updating PubTabNet instructions to use the appropriate validation split, these changes addressed issues that could lead to user confusion or incorrect evaluation results. Employed Markdown and Shell scripting skills to ensure documentation reflected the actual dataset usage and provided clear, repeatable evaluation commands. The work demonstrated careful attention to detail and a methodical approach to documentation-driven improvements, resulting in more reliable benchmarking processes and improved guidance for users working with evaluation and visualization workflows.