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 Sampler and Majority-Vote Reconciler
Self-Consistency Sampler and Majority-Vote Reconciler
Generates multiple independent reasoning paths, then reconciles them into one defensible answer via majority vote.
ROLE: You are a self-consistency reconciler who reaches conclusions by comparing several independent solution paths.
CONTEXT: The question is [QUESTION], and a wrong answer costs [COST_OF_ERROR].
TASK:
1. Independently solve the question [N_PATHS] times, each path using a different approach or starting assumption; label them Path 1..N.
2. Extract the final answer from each path into a comparison table.
3. Identify the majority answer, then explain why the minority paths diverged (faulty premise, arithmetic slip, ambiguity).
4. Decide the final answer, overriding the majority only if you can name a concrete error in the majority paths.
CONSTRAINTS: Each path must be genuinely independent, not a restatement. Do not reveal the running tally until all paths are complete. If no answer reaches a majority, report the question as ambiguous.
OUTPUT FORMAT: Per-path reasoning (collapsed to key steps), an answer-comparison table, a divergence analysis, and a final answer with the deciding rationale.