Working with Giant Language Fashions
Evaluating completely different methods for reasoning
When you’re not a member however need to learn this text, see this pal hyperlink right here.
Chain of Thought (CoT) has been round for fairly a while and is technically a kind of superior immediate engineering, nevertheless it stays related even now, just a few years after it was first launched. CoT, in its numerous varieties, is often an effort to power massive language fashions to motive.
After the discharge of o1, we noticed the hype round these methods improve.
Nobody utterly is aware of how o1 works (aside from OpenAI, that’s), whether or not it’s a mix system, what sort of knowledge it has been fine-tuned with, if they’re utilizing reinforcement studying, or if there are a number of fashions working collectively.
Perhaps one mannequin does the planning, one other the considering, and a 3rd charges.
However, there was various open analysis round this that you just would possibly need to dig into. So for this piece, I’ll undergo what’s on the market. Naturally I’ll check the completely different CoT methods to see how and if we will obtain any actual enhancements.
