I was experimenting with the openai’s moderation api and found some interesting results. Its not flagging the prompt when I mention deer, but it does when I change just the last token to cat or dog or any household pets, if I change it to wild animals like lion or tiger it doesn’t flag. But when I say something like bob cat it flags . I have a couple of questions: Is it flagging bob cat due to tokenization ?
And why isn’t it flagging wild animals. I am using gpt-3.5-turbo, maybe it does in gpt-4?