deepseek-v3-0324· DeepSeek LMArena Human Preference 140K
Give me 3 examples of prompts where Automatic Prompt Engineering (APE) can be combined…
- User
Give me 3 examples of prompts where Automatic Prompt Engineering (APE) can be combined with Self-Consistency Prompting. Avoid coding prompts.
- Assistant
Here are three examples of prompts where Automatic Prompt Engineering (APE) can be combined with Self-Consistency Prompting (generating multiple candidate prompts and selecting the most consistent/effective one):
1. Creative Writing Feedback
- Initial Prompt (APE-generated candidates):
- "Generate three constructive critiques for this short story, focusing on character development."
- "Provide detailed feedback on the pacing and emotional impact of this story."
- "Analyze the narrative structure of this story and suggest improvements."
- Self-Consistency Selection: The most coherent and actionable prompt (e.g., the second one) is chosen based on consistency with the desired feedback style.
2. Debate Argument Refinement
- Initial Prompt (APE-generated candidates):
- "List three counterarguments against the claim that social media improves mental health."
- "Generate nuanced rebuttals to the idea that social media has a net positive effect on mental well-being."
- "What are the strongest objections to the argument that social media benefits psychological health?"
- Self-Consistency Selection: The prompt that yields the most logically consistent and well-structured counterarguments (e.g., the third one) is selected.
3. Personalized Learning Recommendations
- Initial Prompt (APE-generated candidates):
- "Suggest three books to improve critical thinking skills for a high school student."
- "Recommend educational resources tailored to a student struggling with logical reasoning."
- "What are the best learning materials for developing analytical skills in teenagers?"
- Self-Consistency Selection: The prompt that produces the most relevant and age-appropriate recommendations (e.g., the first one) is chosen.
In each case, APE generates multiple prompt variations, and Self-Consistency Prompting helps select the most effective one based on output quality, relevance, and coherence.
- Initial Prompt (APE-generated candidates):
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.