I wonder how much of the value here is from canceling out the positional noise rope produces. I would love to see a table comparing an alibi version of this to

I wonder how much of the value here is from canceling out the positional noise rope produces. I would love to see a table comparing an alibi version of this to an alibi baseline in addition to the rope models here. Crazy gains though congrats to the researchers