Decrypting hidden chain of thought
#Decrypting hidden chain of thought
barren mesa
tardy orbit
Maybe training a tuned lens would already be an improvement?
barren mesa
Maybe training a tuned lens would already be an improvement?
its a small 4-layers model, will tuned lens still help? i am also assuming tuned lens would be too powerful for decrypting ?
tardy orbit
its a small 4-layers model, will tuned lens still help? i am also assuming tuned...
Sorry I thought the point was decrypting. (Don't have a good intuition on if it having just 4 layers you would get very different results)