Multiple methods is a problem-solving technique that tries several approaches, compares their results, and uses agreement to increase confidence. It helps refine uncertain answers but cannot eliminate stochastic model errors. section 9.4.5
A practical workflow combines multiple methods with other prompt-refinement techniques:
- Give explicit, increasingly specific instructions. section 9.4.5
- Try several ways to approach the problem. section 9.4.5
- Ask the model to show its work and add context cues that guide the intended technique. section 9.4.5
Checking a multiplication result
For 234 x 432, three approaches consistently produce 101,088, increasing confidence in that answer. Repeating the exact prompt can still produce incorrect results because language models are stochastic and are not designed for computation. section 9.4.5
