Once all data was ready, I tried training an 8-class model (human + 7 AI models), but since these LLMs are all distilled from similar data, their outputs were too similar. Classification accuracy was only ~50%.
Названа исполнительница роли Наташи Ростовой в «Войне и мире» Андреасяна14:45
，更多细节参见体育直播
В МОК высказались об отстранении израильских и американских спортсменов20:59
Зеленскому стали чаще желать смерти02:42
I hear the objection that someone could use AI to iterate, deliberately making good standalone communication. That’s probably fine. That could be an exception to the “Don’t make me talk to your chatbot” rule. If you keep your reader in mind and put that effort in, you probably aren’t the problem that people complain about. The problem is nonetheless real; some of our social outlets are saturated with it. Try to have some empathy for the frustration.