It seems to me the the results here that ‘instruction tuning strengthens both the use of semantic priors and the capacity to learn input-label mappings, but more of the former’ could be interpreted as some positive evidence for the optimistic case (and perhaps more broadly, for ‘Do What I Mean’ being not-too-hard); summary twitter thread, see especially tweets 4 and 5
It seems to me the the results here that ‘instruction tuning strengthens both the use of semantic priors and the capacity to learn input-label mappings, but more of the former’ could be interpreted as some positive evidence for the optimistic case (and perhaps more broadly, for ‘Do What I Mean’ being not-too-hard); summary twitter thread, see especially tweets 4 and 5