Chain-of-Thought Emergence › Key Parameters
Michael BrenndoerferNovember 7, 20251 min read
Language AI HandbookAll chapters Part 30 of 67Emergent Capabilities
Chain-of-Thought Emergence › Key ParametersOne section · Sections in this chapter9 / 11 ↓
Sections in this chapter11
Reading tools Definitions & inline links
Choose your expertise level to adjust how many terms are explained. Beginners see more tooltips, experts see fewer to maintain reading flow. Hover over underlined terms for instant definitions.
Article links
Make inline references clickable
The key parameters for chain-of-thought prompting and self-consistency are:
- temperature: Controls randomness in generation. Use
temperature=0for deterministic CoT responses, ortemperature=0.7for self-consistency sampling where diversity is needed. - max_tokens: Maximum response length. CoT requires more tokens than direct answers, typically 200-500 tokens for multi-step problems.
- n_samples: Number of reasoning chains for self-consistency. More samples increase accuracy but multiply inference cost linearly.
- model: The language model to use. Larger models (>10B parameters) benefit from CoT while smaller models may not show improvement.