Definition
Plain language
A math tool for picking a set of items that are both individually good and varied as a group, instead of all clustered together.
As stated in the literature
A probabilistic model over subsets that assigns higher likelihood to sets whose elements are both high-quality and mutually dissimilar; used in RHO to select a difficulty-and-diversity-balanced coreset via a single tunable trade-off parameter.
Also called: DPP
Why it matters: It gives a mathematically grounded way to choose samples that are both high-quality and diverse instead of redundant.
For example, when picking photos for an album, it would favor a set that is both sharp and varied rather than ten near-identical shots of the same moment.
Heard on the show
“And the actual selection — picking the ten tasks that best balance hard-and-different — uses a tool called a Determinantal Point Process.”Episode 120 — How an AI Agent Rewrites Its Own Tools, Without an Answer Key