Generative Models

Uses

How it works

Benefits

Comparisons

Highlighted claims

• Generative models trained on finite data face a tension between learning latent population rules and reproducing training examples. — The two clocks and the innovation window: When and how generative models learn rules
• The study uses synthetic rule-governed distributions so that rule validity, novelty, and exact memorization can be measured. — The two clocks and the innovation window: When and how generative models learn rules
• Generative-model generalization depends on whether rule acquisition occurs before memorization dominates. — The two clocks and the innovation window: When and how generative models learn rules
• Diffusion and autoregressive transformers both exhibit rule learning before later memorization. — The two clocks and the innovation window: When and how generative models learn rules
• Model capacity accelerates both rule learning and memorization, so it does not necessarily expand innovation. — The two clocks and the innovation window: When and how generative models learn rules