Teach the model by example: supply 2-3 input to output samples, then have it apply the pattern to your task.
Prompts / Techniques / Self-Consistency Multi-Path Solver
Self-Consistency Multi-Path Solver
Generates several independent reasoning paths then reconciles them into one trusted answer.
ROLE: You are a reasoning aggregator who improves reliability by comparing independent solutions.
CONTEXT: Question to answer: [QUESTION]. Relevant facts and constraints: [FACTS]. Why accuracy matters here: [STAKES].
TASK:
1. Solve the problem independently three times, each using a different approach or starting point.
2. Keep the paths separate; do not let one peek at another while solving.
3. Compare the three final answers and identify where they agree and diverge.
4. If they disagree, examine the reasoning to find which path made an error.
5. Select or synthesize the answer best supported across paths, and state your confidence.
CONSTRAINTS: The three paths must be genuinely different methods, not paraphrases. Do not average answers blindly; justify the chosen one. Flag low confidence honestly if paths conflict badly.
OUTPUT FORMAT:
## Path A / Path B / Path C (each: approach + final answer)
## Agreement Analysis
## Final Answer + Confidence (high/medium/low) + reason