Update README.md

README.md

@@ -38,4 +38,21 @@ If you intend to reproduce the Model A (or a variant), it is critical that you u
 - The dataset was tested using Rank 336 / Alpha 672 fine-tune (the maximum my hardware will reasonably allow)
 - 16K max sequence length is necessary to accomodate each training segment (system prompt + data + model output)
 - Use the Gemma 4 official jinja
-- The model's output is included in the fine-tune (no artificial prompt pairs)
+- The model's output is included in the fine-tune (no artificial prompt pairs)
+
+## Citation
+
+If you use this dataset, please cite it as follows:
+
+```bibtex
+@dataset{your_dataset_key,
+  author       = {Steiniger, Matthew},
+  title        = {Athena-Class Model A: Dataset and Reproducibility Artifacts},
+  month        = September,
+  year         = 2026,
+  publisher    = {Zenodo},
+  version      = {v1.0.0},
+  doi          = {10.5281/zenodo.22861338},
+  url          = {https://doi.org}
+}
+```