Archive: Prompt Engineering
This vault explores techniques for eliciting optimal outputs from LLMs. Moving beyond "magic words," modern prompting is about structuring reasoning traces and providing context-aware demonstrations (Few-Shot).
Advertisement
Technique
Chain of Thought Prompting (CoT)
Wei et al. (2022)
Demonstrates that asking models to "think step by step" dramatically improves performance on complex reasoning tasks (math, logic).
Tree of Thoughts (ToT)
Yao et al. (2023)
A framework that allows models to explore multiple reasoning paths, self-evaluate, and backtrack, mimicking human problem-solving.
Sponsored
2026 Perspective
Automated Prompt Compilers (DSPy 3.0)
Stanford NLP (2025)
The shift from manual prompt engineering to compiled optimization logic for agentic workflows.