Gener Tasks

Abouts

The Gener Tasks currently includes 2 subtasks:

• The gene classification task assesses the model's ability to understand short to medium-length sequences. It includes six different gene types and control samples drawn from non-gene regions, with balanced sampling from six distinct eukaryotic taxonomic groups in RefSeq. The classification goal is to predict the gene type.
• The taxonomic classification task is designed to assess the model's comprehension of longer sequences, which include both gene and predominantly non-gene regions. Samples are similarly balanced and sourced from RefSeq across the same six taxonomic groups, with the objective being to predict the taxonomic group of each sample.

Note: The taxonomic classification dataset is substantial (2GB), which may result in extended training and evaluation time. To accommodate the model's maximum context length, we implement right truncation for sequences that exceed this limit.

How to use

from datasets import load_dataset

# Load gene_classification task
datasets = load_dataset("GenerTeam/gener-tasks",name='gene_classification')

# Load taxonomic_classification task
datasets = load_dataset("GenerTeam/gener-tasks",name='taxonomic_classification')

Citation

@misc{wu2025generator,
    title={GENERator: A Long-Context Generative Genomic Foundation Model},
    author={Wei Wu and Qiuyi Li and Mingyang Li and Kun Fu and Fuli Feng and Jieping Ye and Hui Xiong and Zheng Wang},
    year={2025},
    eprint={2502.07272},
    archivePrefix={arXiv},
    primaryClass={cs.CL},
    url={https://arxiv.org/abs/2502.07272},
}