Old AI models can learn new reasoning tricks without a full rebuild

A new method lets existing large AI models improve their reasoning skills. This means developers can make these models smarter at complex tasks without having to retrain them from scratch.

Training large AI models costs a lot of money and takes a long time. This new method means developers might not have to spend all that time and money just to make an existing model better at reasoning. It could make it easier to add advanced thinking to AI models already in use.