how does finetune work?

[lifei-lifei-lifei]

In the finetune example, tabPFN takes a download data for finetune. the finetuned model must work better at the download data, i'd like to know does the finetuned model work well at the prior trainning data. thanks for your help on my question.