Unlocking reasoning in smaller language models is a great direction

→ View original post on X — @_jasonwei