This is largely a Claude problem and GPT models don't suffer from it to nearly the same extent.