The performance of Marmaragan with GPT-4o on the benchmark is promising, with correct annotations having been generated for 50.7% of the benchmark cases. The results establish a foundation for future work on combining the power of LLMs with the reliability of formal software verification.

That’s basically a coin flip. Even if you made it less successful a user would have higher chances of guessing correctly whether an annotation is true of false.

deleted by creator

some demons from ninth circle of hell

What a rude way to invoke cupsd

A search on Tineye had 0 results

And when it arrives it turns out you need a new computer as well. Also it’s lacking some features you need, and it has some new features nobody asked for. Also fixing it became much harder and doing so can get you to jail.

I don’t know, it sounds like the final thing you hear from an untalented standup comedian before they get booed off the stage on open mic night

Not an expert but I think you’re not invited to meetings about how to stop you from doing more war crimes

Can we please not use the terms ‘conspiracy’ and ‘conspiracy theory’ interchangeably?

Live long and pro spore

And he found out all by himself

N’telligence

Egypt has entered the room

It certainly is one of the posts of all time

I love that his face boobs make him instantly recognizable

It’s an interesting source of data. This post currently has 508 upvotes and 148 downvotes so I guess about 22.5% of lemmy users like anime.

Eh, calling Linus ‘quiet’ kind of cancels it out.

Someone made a mistake? Outrageous.

Nobody was complaining about the English language