Sakana AI introduces QD-Rec, a quality-diversity recommendation system

Sakana AI introduces QD-Rec, a recommendation system that searches over both which options to recommend and the reasons for recommending them.

Instead of returning one best answer, it builds an archive of diverse recommendations, each paired with a natural-language hypothesis about why it fits a decision maker's goals.

Sakana AI says QD-Rec achieved the highest mean LLM-rated quality in most tests on acquisition target discovery, customer targeting and paper recommendation. The company is releasing the code on GitHub so researchers and practitioners can apply the method to their own decision-making tasks.