Sakana AI's QD-Rec applies quality-diversity optimization to open-ended recommendations
SakanaAI · hf · 2026-10-20
Sakana AI proposes QD-Rec, extending open-ended search with foundation models to recommendation for high-stakes decisions. It formalizes open-ended recommendation as building a diverse, high-quality archive where each item carries a natural-language hypothesis for why it fits the decision maker's needs. A contextual bandit selects items while a foundation model generates new hypotheses from archived ones as stepping stones; entries join the archive only if novel or better than their nearest neighbor. Across M&A, customer targeting, and paper/book recommendation, QD-Rec achieves the highest mean quality and builds the largest archive in most settings. Code is open-sourced.
More from Research
- PretrainGym pits frontier agents on pretraining data curation: Kimi K3 wins at 1.042 — xeophon · 2026-10-20
- Claude runs 64-hour autonomous research: 318k protein designs, 17 binders validated — kimmonismus · 2026-10-20
- ACEdit Stabilizes 15-Turn Image Editing With Asymmetric Conditioning — tau · 2026-10-20
- Salesforce Trains Reasoning Models to Batch-Generate Diverse Research Ideas — Salesforce · 2026-10-20
- LightOn's blog details SlideOn's method, results and ablations — MoonL88537 · 2026-10-20
- SlideOn's sparse dims are vocabulary-independent, enabling inverted-index retrieval — tomaarsen · 2026-10-20