Least-to-Most Prompting
Reasoning
Level 2 · Advanced reasoningBreak a complex problem into an ordered sequence of simpler sub-problems, solving each one before moving to the next.
Least-to-most prompting decomposes a complex problem into a sequence of progressively harder sub-problems, then solves them in order — using the answer to each sub-problem as part of the context for the next. This differs from Chain of Thought (which reasons through the whole problem in one pass) and from Step-back (which asks a single broader question first): least-to-most builds a staircase of increasingly specific sub-problems, each one resting on the answer to the previous.
When to use it
Use it for problems that naturally decompose into simpler pieces building on each other — multi-part math problems, tasks with dependent sub-steps, or anything where solving the whole thing at once tends to skip a step.
When to avoid it
Avoid it for problems that don't naturally decompose into an ordered sequence, or where the sub-problems aren't genuinely easier than the whole — forcing an artificial decomposition can hurt more than it helps.
Advantages
- Reduces the chance of skipping a necessary intermediate step
- Makes each part of the reasoning independently checkable
- Particularly effective on problems that are hard mainly because they have many dependent parts, not because any single part is hard
Disadvantages
- Requires the sub-problems to be genuinely ordered and dependent — forcing decomposition where none exists adds overhead without benefit
- More tokens and turns than a single reasoning pass
- If the decomposition itself is wrong, every subsequent step inherits the error
Examples
A train travels 240 km in 3 hours, then increases its speed by 20 km/h for the next 2 hours. What is the total distance travelled? Break this into an ordered sequence of simpler sub-problems, from most foundational to the final goal. Solve each sub-problem in order, explicitly using the answer to each one as part of the context for the next, until you reach the final answer.
Pairs well with
Try it
Fill in the placeholders below and see the compiled prompt update live.
What you want to achieve.
The material to work from.
- 45
- Words
- 279
- Characters
- 0/2
- Placeholders filled
- 1 min
- Reading time
- Completeness0
- Clarity0
- Context0
- Specificity0
- Constraints0
- Format0
- Output0
- Constraints note0
- Success criteria0
- Scope focus100
- Verification step0
- Negative example0
- Confidence0
- Placeholder syntax100
- Markdown structure0
- Fill in the remaining placeholders.
- Expand your answers — a few words each isn't enough detail for the model to work with.
- Add background context so the model understands the situation.