Does o1 engage in counterfeit reasoning and if so can we detect it?

Revisiting Shlomi Sher’s 2023 methodology for identifying counterfeit reasoning

The new strawberry model rolled out in limited edition yesterday. It’s called o1-preview or o1-mini and you can find it in your app or browser if you click on model at the top of the screen here.

This model is their more advanced reasoning model. And Noam Brown at OpenAI, whose Twitter profile says he has been working on reasoning models, including poker…