Update README.md

README.md

@@ -19,7 +19,7 @@ Notable differences with existing architectures include:
 - **Self-attention with 2D Relative Positional Encoding in Polar Coordinates**: This allows the model to effectively process the layout of documents in a polar coordinate system, which is particularly useful for documents with complex layouts (invoices, financial documents, forms, etc.).
 - **No vision feature**: The model does not rely on visual features extracted from images, focusing solely on text and layout information.
 
-For more technical details, see the [DocPolarBERT paper on arXiv](https://arxiv.org/abs/2507.08606).
+For more technical details, see the [DocPolarBERT paper on arXiv](https://arxiv.org/abs/2507.08606) and the [github repo](https://github.com/buthaya/docpolarbert/tree/main).
 
 ## Pre-trained Model