Worked on the PaddlePaddle/Paddle repository to enhance dataset documentation and doctest reliability, focusing on the IMDB, Movielens, and UCI Housing datasets. Used Python and data science expertise to reformat code examples from Python to pycon blocks, improving readability and aligning doctest outputs with expected paddle.Size results. Applied machine learning knowledge to ensure documentation accurately reflected dataset usage, while implementing code-style standardization for consistency. These targeted improvements increased documentation clarity and correctness, helping reduce onboarding time and potential support queries. The work demonstrated attention to detail in documentation and testing, contributing to a more reliable developer experience for dataset users.