HackerTrans
TopNewTrendsCommentsPastAskShowJobs

wjessup

no profile record

Submissions

New Gandalf from Lakera.ai

gandalf.lakera.ai
4 points·by wjessup·3 anni fa·1 comments

GPTGladiator: Make many draft responses, use second model to pick best

github.com
34 points·by wjessup·3 anni fa·5 comments

comments

wjessup
·3 mesi fa·discuss
100% of people who've ever had DHMO have died.

This is scientifically verified and yet nobody does anything about it.
wjessup
·5 mesi fa·discuss
what about the laws that say crossing the border is illegal?
wjessup
·5 mesi fa·discuss
Any commentary about how adversaries won't have regulations?
wjessup
·6 mesi fa·discuss
agreed. this persons world view is skewed, superficial, self-centered.
wjessup
·3 anni fa·discuss
Great basics tutorial: https://colinraffel.com/blog/you-don-t-know-jax.html
wjessup
·3 anni fa·discuss
You forgot jumping right into the derivation first without any context of why it matters.
wjessup
·3 anni fa·discuss
We are experimenting with ways to use ChatGPT to get better answers more reliably, remove hallucinations, etc.

This little library will generate multiple draft responses and then use a second model to judge the answers and pick a winner, which is then returned to the user. Google's Bard uses this same approach.

With this library you can apply the pattern to gpt-3.5 and gpt-4.

Drafts are generated in parallel and all drafts are evaluated with a single prompt.

This will use a lot of tokens. For example to generate 3 drafts, you are at 3x + you need to feed those drafts into another prompt + get that response, so >7x.

Streamlit demo: https://theoremone-gptgladiator-streamlit-ui-5ljwmm.streamli...
wjessup
·3 anni fa·discuss
Thank you for your README, I'm sharing it with my team.