Does anyone have a favorite large language model prompt engineering paper? Especially one that is somewhat applied to a particular task or domain? I’m looking for a good paper to model some current work on.
Thanks!
Conversation
Replying to
You’ve probably already seen this one from couple months ago:
arxiv.org/pdf/2102.07350
2
5
Replying to
I saw it for the first time 5 minutes ago. It’s a bit more abstract that I was going for, since it is about how good prompts can be better than few-shot.
I also had the idea of meta-prompting, but dammit they wrote about that first.
2
2
Laria and Kyle also have a blog: generative.ink that reports a few things beyond their CHI late-breaking paper
1
2
What I really want to know is if anyone has gotten away with writing a paper about how they designed some prompts for GPT3 (or GPT-Neo) that got really good zero-shot performance on a task, beating baselines. What would such a paper look like?
2
Show additional replies, including those that may contain offensive content
Show