Improve attribution: separate weight ownership / model terms / code license

README.md

@@ -10,22 +10,35 @@ tags:
 
 Mirror of the Sei deep-learning model originally hosted at
 `https://zenodo.org/record/4906997/files/sei_model.tar.gz`.
-Maintained by chorus (https://github.com/pinellolab/chorus) for
-resilience — the canonical Zenodo record is durable but single-link;
-chorus mirrors here for redundancy.
+Maintained by [chorus](https://github.com/pinellolab/chorus) for redundancy
+alongside the canonical Zenodo record.
 
 ## Source
 
-- Original: Zenodo record [4906997](https://zenodo.org/record/4906997)
+- Original: Zenodo record [4906997](https://zenodo.org/record/4906997).
 - Paper: Chen, Zhou, Troyanskaya 2022, *A sequence-based global map of
-  regulatory activity for deciphering human genetics*, Nature Genetics.
-- License: CC BY-NC 4.0 (per the original Sei repository).
+  regulatory activity for deciphering human genetics*, Nature Genetics 54,
+  940–949. https://doi.org/10.1038/s41588-022-01102-2
+- Reference repo: https://github.com/FunctionLab/sei-framework
+
+## License & attribution
+
+The weights were copied from the Zenodo record published by the
+Troyanskaya lab. Those weights were created by the Troyanskaya lab and
+are the property of Princeton University / the Sei authors.
+
+The model parameters, outputs, and any derivatives thereof remain subject
+to the **CC BY-NC 4.0** terms applied to the original Sei release.
+
+The model code (the chorus loader at `chorus.oracles.sei`) is released
+under the same chorus license. These licensing terms are consistent with
+the terms for the reference code and the model weights.
 
 ## Contents
 
 `sei_model.tar.gz` — the original Sei tarball, byte-identical to the
 Zenodo record. Extracts to a directory containing the PyTorch model
-weights (.h5) and the chromatin-class label index.
+weights and the chromatin-class label index.
 
 ## Usage