> And yes, the paper is already outdated The paper is about the fact that the benchmark’s evaluator is prone to egregious incorrect rejections of what answers

> And yes, the paper is already outdated The paper is about the fact that the benchmark’s evaluator is prone to egregious incorrect rejections of what answers that it should accept. A new model will not invalidate that issue.