Mele0/SynLV

SynLV v1.0是一个用于评估纵向生存模型在不完整最终遭遇下的鲁棒性的合成基准工具。该版本包括9个主要场景和每个场景的5个队列种子。每个场景/种子目录包含训练、验证和测试集的Parquet文件以及元数据文件。数据集旨在用于在不完整最后访问情况下的鲁棒性基准测试，并使用固定的场景和分割约定进行可重复评估。

SynLV v1.0 is a synthetic benchmark artifact for evaluating robustness of longitudinal survival models under incomplete final encounters. The release includes 9 primary scenarios and 5 cohort seeds per scenario. Each scenario/seed directory contains train, validation, and test sets in Parquet format along with metadata files. The dataset is intended for benchmarking robustness under last-visit incompleteness and reproducible evaluation using fixed scenario and split conventions.