Chain-of-Thought Reasoning Without Prompting

Chain-of-Thought Reasoning Without Prompting is a research paper introducing a method for eliciting step-by-step reasoning from large language models without requiring explicit chain-of-thought prompts. The approach allows models to generate reasoning chains naturally during decoding.