GPT-4 outperforms Med-PaLM 2 with powerful prompting strategies(microsoft.com)

https://www.microsoft.com/en-us/research/blog/the-power-of-prompting/