Prompt Engineered GPT-4 Beats Gemini on all of Google's text benchmarks (opens in new tab)

(microsoft.com)

21 pointsFlyingLawnmower2y ago3 comments

3 comments

"We note that Medprompt+ relies on accessing confidence scores (logprobs) from GPT-4. These are not publicly available via the current API but will be enabled for all in the near future."

Had that been announced yet?

alecco2y ago

It would be interesting to see how this can be used for open models.

byyoung32y ago

engineered? Over overfit?

j / k navigate · click thread line to collapse