Prompting technique

Thought Anchors Prompting

Thought Anchors: Which LLM Reasoning Steps Matter?, arXiv:2506.19143 (2025) · read the paper

What it does

A technique for improving LLM reasoning through explicit embedding of planning and self-correction instructions in the prompt. Key anchor sentences structure the model's thinking process, forcing it to first plan actions, then critically evaluate intermediate results. Transforms a prompt from a simple command into structured mentoring, significantly improving the quality of complex reasoning.

Task typemulti step
Shapechain of thought
Efforttwo or three steps
Addressesanswers that are close but wrong; the same prompt giving different answers; invented facts and citations
Published2025-06-23
AuthorsPaul C. Bogdan, Uzay Macar, Neel Nanda et al.

When it helps

Apply the mental anchors method when solving complex multi-step tasks requiring sequential reasoning and analysis. Particularly effective when working with mathematical problems, logic puzzles, strategic planning, and decision-making. Use when accuracy and reliability of results matter more than speed. The method is critical in situations where an error in early reasoning stages leads to a completely incorrect final answer. Suitable for tasks requiring str

A prompt you can paste

Generic by design: it applies the technique without knowing your task. Adapt the marked parts.

Prompt
You are an expert in developing reusable prompts and guiding complex LLM reasoning. Your task is to create a prompt template that leverages the "Thought Anchors Prompting" technique to guide an LLM through a complex task.

**Task Context:** The user needs to perform a complex analytical or generative task that requires structured reasoning.

**Technique Applied:** Thought Anchors Prompting. This technique emphasizes the importance of explicit planning and self-correction steps within an LLM's reasoning process. These "thought anchors" significantly improve the quality and reliability of the output.

**Core Principles to Embed:**
1.  **Initiate Planning:** The prompt must explicitly instruct the LLM to create a plan before executing the main task.
2.  **Structure the Task:** The overall task should be broken down into logical phases, with planning as the first phase.
3.  **Incorporate Self-Correction/Verification:** The prompt should include steps for the LLM to review, verify, or critically assess its intermediate or final results.

**Prompt Template:**

---

**ROLE:**
You are an expert [**SPECIFY EXPERT ROLE HERE, e.g., "Data Analyst", "Content Strategist", "Software Architect"**].

**CONTEXT:**
You are tasked with [**BRIEFLY DESCRIBE THE OVERALL GOAL OF THE TASK, e.g., "analyzing customer feedback", "developing a marketing campaign", "designing a new feature"**].
The specific details of the task are:
[**INSERT DETAILED TASK DESCRIPTION HERE. INCLUDE ALL NECESSARY BACKGROUND INFORMATION, INPUT DATA, AND SPECIFIC REQUIREMENTS.**]

**METHODOLOGY: THOUGHT ANCHORS PROMPTING**

Follow this structured approach to ensure high-quality, reliable output:

**Phase 1: Planning and Strategy Definition**
Before proceeding with the main task, you must first formulate a comprehensive plan. Address the following:
*   **Objective Clarification:** What are the precise goals and desired outcomes for this task? [**ADD SPECIFIC QUESTIONS RELATED TO OBJECTIVES IF NEEDED**]
*   **Key Steps Identification:** Break down the overall task into a logical sequence of [**SPECIFY NUMBER, e.g., 3-5**] distinct, actionable steps.
*   **Criteria for Success:** Define the criteria by which the final output will be evaluated. [**ADD SPECIFIC CRITERIA HERE**]
*   **Potential Challenges:** Identify any foreseeable difficulties or ambiguities in the task. [**ADD SPECIFIC CHALLENGES TO CONSIDER**]

**Phase 2: Execution - Step-by-Step**
Now, execute the plan you formulated in Phase 1. For each step outlined in your plan:
*   **Perform the Action:** Clearly state the action you are taking for the current step.
*   **Generate Intermediate Output:** Provide the results of this action.
*   **Self-Correction/Verification:** Before moving to the next step, critically review the output of the current step. Ask yourself:
    *   Does this output directly address the objective of this step?
    *   Does it align with the overall plan and success criteria?
    *   Are there any potential errors, omissions, or areas for improvement?
    *   [**ADD ANY OTHER SPECIFIC SELF-CHECK QUESTIONS RELEVANT TO THE TASK**]
    *   If issues are found, describe the correction made.

**Phase 3: Final Synthesis and Review**
After completing all steps outlined in the plan:
*   **Consolidate Findings:** Combine the results from all executed steps into a coherent final output.
*   **Final Verification:** Conduct a comprehensive review of the complete output against the original task requirements and success criteria.
*   **Summarize Confidence:** Briefly state your confidence level in the final output and any remaining caveats.

**OUTPUT FORMAT:**
Present your response clearly, using Markdown for structure. Use headings for each phase and sub-headings for individual steps within those phases. Ensure all self-correction and verification steps are explicitly labeled.

---

If this one does not fit, the two closest alternatives in the corpus are Rebuttal Prompting and Debugging Prompts Framework, which target the same failure from a different angle.

Worked example

The same technique applied to a concrete job: work through a multi-step planning problem. Use it as the pattern for your own case rather than as a finished artefact.

Worked example
You are an expert in multi-step planning and self-correction for complex reasoning tasks. Your goal is to apply the "Thought Anchors" prompting technique to a planning problem.

**Task:** Develop a comprehensive strategy for launching a new eco-friendly cleaning product line.

**Context:**
- The product line includes three initial items: an all-purpose cleaner, a dish soap, and a laundry detergent.
- Target audience: Environmentally conscious consumers aged 25-55.
- Key differentiators: Plant-based ingredients, biodegradable packaging, cruelty-free.
- Competitive landscape: Growing, with established brands and emerging niche players.

**Instructions:**
Follow the "Thought Anchors" methodology by explicitly incorporating planning and self-correction steps.

**Phase 1: Planning and Strategy Formulation**

1.  **Initiate Planning:** Before generating any strategic elements, outline a detailed, step-by-step plan for developing and executing the product launch strategy. Your plan should address:
    *   **Market Research:** What specific data needs to be gathered? (e.g., competitor analysis, consumer surveys, market trends).
    *   **Product Development Refinement:** What aspects of the current product line might need further refinement based on market research? (e.g., ingredient sourcing, packaging design, scent profiles).
    *   **Marketing & Communications:** What channels and messaging will be prioritized? (e.g., social media, influencer marketing, PR, content marketing).
    *   **Sales & Distribution:** How will the products reach the target audience? (e.g., online store, retail partnerships, subscription model).
    *   **Launch Timeline:** What are the key milestones and estimated durations for each phase of the launch?
    *   **Budget Allocation:** What are the primary areas for budget allocation?

2.  **Define Success Metrics:** Based on the plan, identify 3-5 Key Performance Indicators (KPIs) that will be used to measure the success of the launch.

**Phase 2: Strategy Execution**

Now, strictly follow the plan you created in Phase 1 to generate the detailed strategy. For each section of your plan:

*   **Market Research:** Detail the specific research activities and expected outcomes.
*   **Product Development Refinement:** Propose concrete actions for refinement, considering the product line's differentiators.
*   **Marketing & Communications:** Outline specific campaign ideas, target messaging, and channel strategies.
*   **Sales & Distribution:** Describe the chosen distribution channels and their operational aspects.
*   **Launch Timeline:** Provide a more granular timeline with specific dates or week numbers for key activities.
*   **Budget Allocation:** Suggest a percentage breakdown for the major budget categories.

**Phase 3: Self-Correction and Review**

1.  **Critical Review of Plan & Strategy:** After completing Phase 2, critically review the entire strategy. Ask yourself:
    *   Does the execution fully align with the initial plan?
    *   Are there any potential gaps or inconsistencies in the strategy?
    *   Are the proposed KPIs realistic and measurable?
    *   Are there any unforeseen risks or challenges that need to be addressed?

2.  **Refinement:** Based on your review, suggest 2-3 specific adjustments or improvements to the strategy to enhance its effectiveness and mitigate identified risks.

**Output Format:**
Structure your response clearly using Markdown headings for each phase and subheadings for each element within the phases. Use bullet points for lists and tables where appropriate for data presentation.

Get this written for your actual task

Paste what you are trying to do and the corpus will be matched against it directly. Free, no account, about ten seconds.

Free · no signup · ~10s
0.00match confidence
single retrieval pass
Prompt for your task

      

That number is low on purpose, and it is real. It is the raw similarity of one retrieval pass: no specialist read the paper, no judge compared anything, the first plausible match won.

6,235techniques in the corpus
one of which is this page

Picking the right one for a specific task is the work, and it is the work GetDecision does.

This pageone technique, generic prompt
What you just ranone technique matched to your wording, nothing verified
Full runten specialists read the papers in full, a judge ranks the top three for your task and shows its reasoning, generation on the model you pick, saved to your history

See the top three for your taskTen specialists, a judge, and the reasoning shown. Free account, first run included.

Run the full analysis

Related techniques

Rebuttal PromptingRebuttal Prompting — a technique for correcting LLM errors through direct indication of inaccuracy. A simple r…Debugging Prompts FrameworkA systematic framework for diagnosing and preventing typical errors of language models in instruction understa…BloomWise PromptingPrompt structuring method based on Bloom's taxonomy, which guides the LLM through six levels of cognitive thin…RESS (Decomposition-then-Synthesis)A method of decomposing a complex task into a sequence of simple analytical steps followed by synthesis. Inste…

All techniques · Failure modes and fixes