← back
Prompt Engineering Tactics: Dan Cleary
Takeaway
Three cheap, research-backed prompt edits—multi-persona, 'according to', and emotional stimuli—measurably lift output quality with no code change.
Summary
- Dan Cleary (PromptHub) shares three research-backed prompt tactics for everyday and product use.
- Multi-persona prompting (UIUC): instruct the model to spawn task-specific personas (publicist, author, target audience) that brainstorm before producing output—great for generative tasks.
- 'According to <source>' grounding (Johns Hopkins) reduces hallucinations up to 20% by anchoring retrieval to a named source.
- EmotionPrompt (Microsoft et al.): adding emotional stimuli like 'this is important for my career' to the end of prompts yields 8-115% gains depending on task.
prompt-engineeringtacticshallucinations
Original description
Everything is going well until a prompt that works most of the time, goes off the rails. Does this sound familiar? The inherent fragility and non-deterministic nature of prompts can be challenging, often leading to inconsistent outputs. In this session, I'll dive into advanced prompting techniques that are grounded in research and empirical evidence. We'll explore how to tailor methods to specific tasks and how these strategies can significantly enhance output consistency and quality. Recorded & streamed live for the AI Engineer Summit 2023. See the full schedule of talks at https://ai.engineer/summit/schedule & join us at the AI Engineer World's Fair in 2024! Get your tickets today at https://ai.engineer/worlds-fair About Dan I'm the co-founder of PromptHub, a platform that makes it easy to iterate, test, and manage prompts. I'm a runner, basketball junkie, and Hubermanlab listener.