The Hindsight Neglect Problem

In terms of like literal and it's not just all this seems really good. But there's another test that they did in this paper. It's actually called the Hindsight neglect problem, which is a decision-making Exercise. GPT-3.5 go to score near Let's say 20% mm-hmm. So it goes 0 to 100 and then guess what GPT-4 was 100 Wow perfect score so it's like that's the difference we're talking about. In terms of reasoning there's some Kind of what it was was some human test that humans do as well and humans No, no, no,no, it can do that as well but that's not