4 ms·

I am not up-to-speed on CoT side but is this similar to how perplexity does it ie. - generate a plan - execute the steps in plan (search internet, program this

by asadm 2y ago

I am not up-to-speed on CoT side but is this similar to how perplexity does it ie. - generate a plan - execute the steps in plan (search internet, program this part, see if it is compilable) each step is a separate gpt inference with added context from previous steps. is O1 same? or does it do all this in a single inference run?

seydor 2y ago
that is the summary of the task it presents to the user. The full chain of thought seems more mechanistic
golol 2y ago
There is a hige difference which is that thex use reinforcement learning to make the model use the Chain-of-Thought better.