Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

Why this source is here

Established that intermediate reasoning steps at inference change measured performance.

Verification

Content-verified on 2026-07-27: the canonical source and its title were resolved during the Atlas review. This is not an endorsement of the source’s argument.

Authors

Wei et al.

Identifier

arXiv:2201.11903