[synarchefriend]

o1 can also solve many arbitrary math problems that it could not have possibly seen in its training data. And it shows the steps that it uses to do so. How do you explain this without reasoning?