I noticed the top_logprobs returned by the content filter endpoint returns two additional labels not mentioned in the documentation, "3" and " 2". Does anyone know the uses for these labels? These additional labels seem to be something different based on the prompt. Sometimes it’s a word or a letter.