QD-Rec explores different reasons to recommend items from a fixed pool

Sakana AI describes QD-Rec as a way to build a diverse collection of recommendations, each linking an existing option to a reason it may fit a decision maker. In tests involving acquisition targets, customer targeting and paper recommendations, the team reports the highest mean LLM-rated quality in most settings against simpler variants. Sakana AI has released the code and says practical use will need up-to-date information and expert human feedback.