flashfaffe2
Joined 5 years ago
Karma
167
Submissions
57
Member since
Oct 8
Joined 5 years ago
The post seems to propose the following vector:
Idea-> LLm validation -> confidence -> no further checks
My process is more :
Idea-> LLm response -> skeptical reflection -> adversarial prompting -> synthesis
"Bandits work best for relatively simple optimization choices at very large scale"
Have you considered different methods to address this shortcoming?
Thanks for sharing the article