Update README.md

README.md

@@ -22,7 +22,7 @@ The training and evaluation code is available at the official [TripPy repository
 
 The model was trained on MultiWOZ 2.1 data via supervised learning using the [TripPy codebase](https://gitlab.cs.uni-duesseldorf.de/general/dsml/trippy-public).
 MultiWOZ 2.1 data was loaded via ConvLab-3's unified data format dataloader.
-The pre-trained encoder is [RoBERTa](https://arxiv.org/abs/1907.11692) (base).
+The pre-trained encoder is [RoBERTa](https://huggingface.co/docs/transformers/model_doc/roberta) (base).
 Fine-tuning the encoder and training the DST specific classification heads was conducted for 10 epochs.
 
 ### Training hyperparameters