Update README.md

README.md

@@ -39,10 +39,14 @@ The dataset used for training is listed below according to training sequences:
 1. [MRQA(train split)](https://huggingface.co/datasets/mrqa)
 2. [UCLNLP/adversarial_qa](https://huggingface.co/datasets/UCLNLP/adversarial_qa)
 3. [mbartolo/synQA](https://huggingface.co/datasets/mbartolo/synQA)
-4. [MRQA(test split)](https://huggingface.co/datasets/mrqa)
+4. [MRQA(test split)](https://huggingface.co/datasets/mrqa)*This model
 
 ## Training procedure
 
+The training approach uses the fine-tuning approach of transfer learning on the pre-trained model to perform NLP QA tasks. 
+Each time a model was trained with one dataset only and saved as the PTMs for the next training. 
+This model is the last model that trained with [MRQA(test split)](https://huggingface.co/datasets/mrqa). 
+
 ### Training hyperparameters
 
 The following hyperparameters were used during training: