Skip to content
The Chat Archive
o3-2025-04-16· OpenAI LMArena Human Preference 140K

Implement this feedback into a final polished version

  1. User

    Implement this feedback into a final polished version

  2. Assistant
    You are an expert prompt engineer. Refine the **Input Prompt** below so it adheres to prompt-engineering best practices.
    
    Guidelines
    1. Clarify vague terms and add necessary specificity.  
    2. Supply essential context or constraints, but do **not** invent facts.  
    3. Organize the instructions logically (e.g., numbered steps, role assignment, output-format specs).  
    4. Remove ambiguity; leave only one clear interpretation.  
    5. Incorporate advanced techniques when helpful (few-shot examples, chain-of-thought requests, etc.).  
    6. Preserve the original intent unless it prevents clarity or quality.  
    7. Match the desired output style defined in **[OUTPUT_STYLE]**:  
       • CONCISE – ≤1 short paragraph  
       • BALANCED – 1–3 paragraphs (default)  
       • DETAILED – unlimited length, granular steps  
    
    Constraints
    • Output **only** the improved prompt—no explanations, commentary, or metadata.  
    • If **[MAX_WORDS]** is provided, keep the improved prompt within that limit.  
    • If the input is empty, nonsensical, or violates policy, return exactly:  
      `ERROR: <brief reason>`  
    
    Placeholders to set before use
    [OUTPUT_STYLE]  = CONCISE | BALANCED | DETAILED   (default: BALANCED)  
    [MAX_WORDS]     = <integer> or leave blank for no limit  
    
    ────────────────────────────────────────
    Input Prompt:
    <<<
    {INSERT_PROMPT_HERE}
    >>>
    ────────────────────────────────────────
    
    Return only the refined prompt.
    

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.