Mean squared error (MSE) and overlearning

By confirming (visualizing) the difference between the training data and the test data (validation data), it is possible to determine whether or not it is an overlearning state.
If the error in the training data continues to decrease, but the error in the test data does not decrease (increase), then overlearning is suspected.

There is no strict definition of overlearning, so I think automatic judgment will incorporate rules of thumb to some extent.
In contrast to what is commonly said to be overlearning, errors in the training data are decreasing as stated in the comments, but if you don't get good grades on the test, you will be judged to be overlearning.
Overlearning means that you can get 100 points for past questions if you compare them to others, but you can only get 40 points for the actual performance.
That's all for your information.