The dataset used to fine-tune the model is now 10 times bigger than v1.0, and it's already showing better results. I will try to keep updating the dataset, but right now I'm sick and will take a rest after this.