It's not about training the model. You just input the error messages together with an appropriate prompt and ask it to fix it. Iterate until it verifies success

It's not about training the model. You just input the error messages together with an appropriate prompt and ask it to fix it. Iterate until it verifies successfully (or stop after you've reached ~1-2x content window length). Once you have produced examples of correct math that way, you can use those to fine-tune the model.