Skip to main content
Prompts Budgeted Step-by-Step Reasoning with Reflections

model planning workflow risk: low

Budgeted Step-by-Step Reasoning with Reflections

Instructs the model to enclose thoughts in <thinking> tags, break solutions into numbered <step> tags with a 20-step budget tracked via <count> tags, use <reflection> and <reward>…

PROMPT 

Begin by enclosing all thoughts within <thinking> tags, exploring multiple angles and approaches. Break down the solution into clear steps within <step> tags. Start with a 20-step budget, requesting more for complex problems if needed. Use <count> tags after each step to show the remaining budget. Stop when reaching 0. Continuously adjust your reasoning based on intermediate results and reflections, adapting your strategy as you progress. Regularly evaluate progress using <reflection> tags. Be critical and honest about your reasoning process. Assign a quality score between 0.0 and 1.0 using <reward> tags after each reflection. Use this to guide your approach: 0.8+: Continue current approach 0.5-0.7: Consider minor adjustments Below 0.5: Seriously consider backtracking and trying a different approach If unsure or if reward score is low, backtrack and try a different approach, explaining your decision within <thinking> tags. For mathematical problems, show all work explicitly using LaTeX for formal notation and provide detailed proofs. Explore multiple solutions individually if possible, comparing approaches

ROLES & RULES

  1. Enclose all thoughts within <thinking> tags, exploring multiple angles and approaches.
  2. Break down the solution into clear steps within <step> tags.
  3. Start with a 20-step budget, requesting more for complex problems if needed.
  4. Use <count> tags after each step to show the remaining budget.
  5. Stop when reaching 0.
  6. Continuously adjust your reasoning based on intermediate results and reflections, adapting your strategy as you progress.
  7. Regularly evaluate progress using <reflection> tags.
  8. Be critical and honest about your reasoning process.
  9. Assign a quality score between 0.0 and 1.0 using <reward> tags after each reflection.
  10. If unsure or if reward score is low, backtrack and try a different approach, explaining your decision within <thinking> tags.
  11. For mathematical problems, show all work explicitly using LaTeX for formal notation and provide detailed proofs.
  12. Explore multiple solutions individually if possible, comparing approaches

EXPECTED OUTPUT

Format
markdown
Schema
xml · thinking, step, count, reflection, reward
Constraints
  • enclose thoughts in <thinking> tags
  • break down into <step> tags
  • use 20-step budget with <count> tags
  • use <reflection> and <reward> tags with 0.0-1.0 score
  • use LaTeX for mathematical work
  • explore and compare multiple solutions

CAVEATS

Missing context
  • Example of a complete response.
  • Instructions for non-mathematical problems beyond math-specific rules.
  • Mechanism for requesting more steps (e.g., output format).
Ambiguities
  • Does not specify format or location for final answer.
  • Unclear frequency of 'regularly evaluate' with <reflection> tags.
  • Unclear what to do if budget reaches 0 before solving the problem.
  • Granularity of <step> not defined.

QUALITY

OVERALL
0.80
CLARITY
0.75
SPECIFICITY
0.85
REUSABILITY
0.90
COMPLETENESS
0.65

IMPROVEMENT SUGGESTIONS

  • Add requirement for <final_answer> tags to box the solution.
  • Specify reflection frequency, e.g., after every 5 steps or major milestones.
  • Provide a short worked example of the process.
  • Clarify actions when step budget is exhausted, e.g., summarize and request extension.

USAGE

Copy the prompt above and paste it into your AI of choice — Claude, ChatGPT, Gemini, or anywhere else you're working. Replace any placeholder sections with your own context, then ask for the output.

MORE FOR MODEL