Prompt workshop
What is GEPA?
GEPA is reflective prompt evolution with Pareto selection: seed, execute, reflect, mutate, keep the non-dominated archive.
It is also a DSPy optimizer. Related research includes SePO and MASPO, filed here as library system prompts that describe similar loops.
Why it matters
GEPA (Generative Evolution with Pareto Archive) evolves prompts through reflection and mutation, keeping only the non-dominated solutions. It combines the search power of evolution with the interpretability of textual feedback.
GEPA is also available as a DSPy optimizer, making it accessible without implementing the algorithm from scratch.
Key takeaways
- 1GEPA evolves prompts through reflection, mutation, and Pareto selection.
- 2It keeps an archive of non-dominated solutions.
- 3Available as a DSPy optimizer for easy integration.
Common mistakes
- ✕Running GEPA without a clear metric — evolution needs a fitness function.
- ✕Not inspecting the evolved prompts to ensure they are coherent.
Related terms
Concept neighborhood
Terms linked from GEPA in the glossary graph.