ViroHyena

ViroHyena is a Hyena-based nucleotide language model pre-trained on the ViroBlend corpus, a 216 Mbp mixed pretraining dataset with source-wise stratified sampling to balance human reference, multi-species genomes, and viral in-domain sequences.

The model was introduced as a baseline in the paper ViroBench: Benchmarking Nucleotide Foundation Models on Viral Genomics Tasks.

• Repository: https://github.com/QIANJINYDX/ViroBench
• Project Page: https://sii-agi4s.github.io/ViroBench/

Model Configurations

ViroBench

ViroBench is a comprehensive diagnostic benchmark for viral sequence modeling, designed to evaluate models across two critical dimensions:

• Understanding Axis: Taxonomy classification and host prediction.
• Biosecurity Axis: Genome modeling and CDS completion (risk-aware generation diagnostics).

Citation

@article{ye2026virobench,
  title={ViroBench: Benchmarking Nucleotide Foundation Models on Viral Genomics Tasks},
  author={Ye, Dongxin and Hu, Fang and Hu, Han and Hu, Shu and Tan, Yang and Ouyang, Wanli and Li, Stan Z and Cui, Jie and Dong, Nanqing},
  journal={arXiv preprint arXiv:2605.25388},
  year={2026}
}