All templates
⚡ Advanced Techniques·Works on: claude, chatgpt, gemini
Self-Consistency Voting
Technique: Self-Consistency
Reason the same question 5 different ways. Take the majority answer. Beats single-sample CoT.
Advanced#self-consistency#voting#reasoning
0/1
Fill the template
1 placeholder left.
~203tokens
Live preview· click any pill to jump to its field
Reason about this problem 5 INDEPENDENT times. Treat each attempt as if you'd never seen the problem before. Problem: Protocol: For each of attempts 1–5: - Restate the problem in your own words. - Solve it step by step, using a DIFFERENT reasoning approach where possible (algebra vs. enumeration vs. estimation vs. constraints-first vs. brute-force). - State the final answer. After all 5 attempts: - Tally the final answers. - The majority answer is the candidate winner. - If 3+ attempts agree, confidence = HIGH. - If 2/2/1 split, confidence = MEDIUM and explain the disagreement. - If 5 different answers, confidence = LOW and recommend the user double-check externally. Return the winning answer + confidence + a 1-line explanation of why the dissenting attempts disagreed.
Inputs0 of 1
›See the lazy version this template replaces
Before — the lazy prompt
Solve <problem>.
Why it works
- Single CoT chains can be confidently wrong; majority voting across diverse approaches catches the brittle ones.
- Forcing different reasoning approaches per attempt — algebra vs enumeration vs estimation — surfaces approach-specific blind spots.
- Explicit confidence calibration based on agreement is the move most prompts skip.
Make this one yours
Replace the bracketed placeholders, then paste into the Prompt Fixer to lint your customisation before hitting send.