4 ms·
Computers have historically been good at answering word problems fed to them verbatim?
by CamperBob2 2mo ago
Computers have historically been good at answering word problems fed to them verbatim?
- ACCount37 2mo agoNo, but they were good at answering formalized versions of the same word problems. What this tells us is that a 3B LLM can retain enough NLU to understand those word problems. Which isn't particularly surprising? And also that the same LLM can solve a math or logic problem it understands. Which is a lot more impressive, because early LLMs were already quite good at NLU, but notoriously bad at things like math, logic and iterative problem solving. This 3B model existing tells us we're beginning to figure out how to imbue models with those capabilities reliably.
- CamperBob2 2mo ago"Formalizing the problem" is pretty much the whole shooting match. It takes intelligence to do that. The rest is mere calculation.