Update README.md

README.md

@@ -38,7 +38,7 @@ widget:
 
 ## Model Description
 
-**HViLM (Human Virome Language Model)** is a genomic foundation model adapted to virus sequences through continued pre-training of [DNABERT-2](https://huggingface.co/zhihan1996/DNABERT-2-117M). HViLM-base was trained on approximately **5 million non-redundant virus-derived sequence fragments** from the [VIRION](https://virion.verena.org/) resource, representing approximately 9,000 virus species across 45+ families.
+**HViLM (Human Virome Language Model)** is a genomic foundation model adapted to virus sequences through continued pre-training of [DNABERT-2](https://huggingface.co/zhihan1996/DNABERT-2-117M). HViLM-base was trained on approximately **5 million non-redundant virus-derived sequence fragments** from the from [VIRION](https://www.viralemergence.org/virion) ([Zenodo deposit](https://doi.org/10.5281/zenodo.15643003)), representing approximately 9,000 virus species across 45+ families.
 
 The architecture and tokenizer remain those of DNABERT-2; continued pre-training updates the model weights using a masked-language-modeling objective on the virus-focused corpus.
 
@@ -331,7 +331,7 @@ If you use DNABERT-2 directly or build on its architecture, please also cite the
 
 ## Acknowledgments
 
-HViLM builds on DNABERT-2 by Zhou et al. Continued-pretraining data were derived from the VIRION resource maintained by the Viral Emergence Research Initiative (Verena).
+HViLM builds on DNABERT-2 by Zhou et al. Continued-pretraining data were derived from the from [VIRION](https://www.viralemergence.org/virion) ([Zenodo deposit](https://doi.org/10.5281/zenodo.15643003)) maintained by the Viral Emergence Research Initiative (Verena).
 
 ---