Qiuzesong focused on enhancing the reliability of the ModelTC/LightX2V repository by addressing reproducibility issues in the WanStepDistillScheduler’s noise generation process. Using Python and deep learning techniques, Qiuzesong implemented a deterministic noise pipeline that leverages a specified generator and fixed shape, ensuring consistent noise output across experimental runs. This technical approach improved the reproducibility of experiments, increased benchmarking reliability, and supported ongoing model development efforts. The work prioritized stability and traceability, introducing minimal changes to reduce regression risk. Although no new features were released, the depth of the solution reflects a strong commitment to robust, reproducible machine learning workflows.