Python Neural Networks Overfitting

I have three separate neural networks (LSTM w/ attention, CNN-LSTM, CNN-LSTM w/ attention), but each of the models seems to overfit or accuracy fluctuates between 82 and 83%. I am not sure whether it is an issue with the model or the dataset. I have tried to add regularization to the model but did not seem to help.

I tried to clean it in different ways, and it does seem to affect the accuracy.

Accuracy right now was 81-82% now.

The accuracy depended on which version of the cleaned dataset I used

you can ignore the lstm model. i replaced the lstm with bert.

I need to get 85% + accurancey for cnn-lstm, lstm-attention, and cnn-lstm-attention only.

thank you.