Least-to-Most Prompting
A technique that decomposes complex problems into simpler subproblems, solving them in order from easiest to hardest.
Also known as: Progressive Prompting, Decomposition Prompting, Subproblem Solving
Category: Techniques
Tags: ai, prompting, reasoning, llm-techniques, problem-solving, decomposition
Explanation
Least-to-Most Prompting is a problem-solving technique that breaks down complex problems into a sequence of simpler subproblems, solving them progressively from easiest to hardest. Each solution builds context for the next, enabling the model to tackle problems that exceed its single-step capabilities.
The two-phase approach:
**Phase 1 - Decomposition**:
Prompt: "To solve [complex problem], what simpler problems do I need to solve first?"
Output: List of ordered subproblems from simplest to most complex
**Phase 2 - Sequential Solving**:
For each subproblem (easiest first):
- Solve using previous solutions as context
- Add solution to accumulated context
- Proceed to next subproblem
Example for "How many tennis balls fit in a school bus?":
1. What are the dimensions of a tennis ball? → ~6.7cm diameter
2. What are the interior dimensions of a school bus? → ~2.4m × 2m × 12m
3. What's the volume of the bus interior? → ~57.6 cubic meters
4. Accounting for packing efficiency (~64%), how many balls fit? → ~500,000
Key benefits:
- Enables solving problems beyond single-prompt capability
- Each step is verifiable independently
- Builds understanding progressively
- Works well with compositional generalization
Least-to-Most is particularly effective for:
- Mathematical word problems
- Multi-step reasoning tasks
- Symbolic reasoning
- Tasks requiring compositional reasoning
Related Concepts
← Back to all concepts