Definition
Plain language
A search method that keeps a whole grid of good-but-different solutions instead of chasing one single best answer.
As stated in the literature
A quality-diversity evolutionary algorithm that maintains an archive of high-performing solutions across a grid of behavioral descriptors; cited as one of the framework-specific search machineries that optimize-anything's declarative loop sidesteps.
Why it matters: It preserves a diverse set of strong solutions, which helps avoid getting stuck on one narrow answer and offers options suited to different conditions.
For example, rather than returning only the single fastest robot gait, it keeps a collection of gaits that are each best for a different body size or step length.
Heard on the show
“You don't need to be an expert in island topologies or MAP-Elites or any of the framework-specific machinery.”Episode 065 — One Loop to Optimize Them All: A Universal API for LLM-Driven Discovery