Collected sources and patterns will appear here. Add from search, explore, or the patterns library.
RawCoT -> RefinedCoT
Refine a chain-of-thought rationale iteratively by prompting a model to critique and correct its previous reasoning steps.
Problem it solves
Single-pass chain-of-thought reasoning is prone to logical leaps, hallucinated facts, and math errors.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.