3 ms·

I think that's a prime example showing that token prediction simply isn't good enough for correctness. It never will be. LLMs are not designed to reason about c

by beders 1y ago

I think that's a prime example showing that token prediction simply isn't good enough for correctness. It never will be. LLMs are not designed to reason about code.