ModelRefs / Self-Consistency — Prompt Pattern
Self-Consistency — Prompt Pattern
Sample multiple chain-of-thought traces and pick the modal final answer. Lifts accuracy on reasoning benchmarks at N× cost.
Overview
Run the same CoT prompt N times with temperature > 0, parse all final answers, and return the most common one. Lifts accuracy on reasoning benchmarks at N× cost.
When to use it: You already use chain-of-thought and need higher accuracy on a closed-form answer.
Pattern details
| Pattern class | reasoning |
|---|---|
| Difficulty | intermediate |
| Also known as | majority voting, ensemble cot |
| Last reviewed | 2026-06-07 |
Known failure modes
- Cost blowup — 5× tokens per query. Mitigation: Cache the base prompt; only sample on hard items.
- Tied modes — Two answers tie; default tie-break may be wrong. Mitigation: Add a verifier judge prompt for ties.
When not to use it
- Sampling open-ended generation (no canonical answer to vote on).
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Self-Consistency — Prompt Pattern.
Frequently asked questions
When should I use the Self-Consistency prompt pattern?
You already use chain-of-thought and need higher accuracy on a closed-form answer.
What are common failure modes of Self-Consistency?
Cost blowup • Tied modes
Does Self-Consistency work across models?
Yes when the anatomy is preserved — see model compatibility matrix on the pattern page.