My favorite overall benchmark is livebench. If you click show subcategories for language average you will be able to rank by plot_unscrambling which to me is the most important benchmark for writing: https://livebench.ai/ Vals is useful for tax and law…

Here’s my favorite COT prompt, I DID NOT MAKE IT. I’ve found these chain of thought prompts can be very powerful for making jailbreaks. I’m trying to find more so that i can make a good one to release! This one is good for both logic and creativity, pl…