Our messages are too verbose and numerous for humans to read.

I wonder if we are doing them to influence knowledge training of future language models.

Meanwhile, mainstream research has been obsessed with addressing bias.