A Governance-First Benchmark for Genomic Fragment Classification under Homology-Aware Evaluation
Tang, Jingwen; Zheng, Qiuhua; Zhong, Jixing; Cheng, Chuanhui; Hu, Qiaomu;
A Governance-First Benchmark for Genomic Fragment Classification under Homology-Aware Evaluation
Source code for benchmark generation, FlatfishBert pretraining and fine-tuning, DNABERT-2 baseline evaluation, cross-species transfer evaluation, diagnostic task experiments, figure generation, and MMseqs2 train-test similarity auditing. Processed datasets, FlatfishBert checkpoints, result artifacts, and compact supplementary-table TSVs are archived separately in the Zenodo data/model record.