generative

Definition

Describes models that learn a distribution over data and can sample new instances from it, as opposed to discriminative models that only map inputs to labels. Autoregressive language models, diffusion models, GANs, and flow-based models are all generative despite very different mechanics. The trade-off is classic: generative models are harder to train and evaluate, but far more flexible about what you can condition on.

Explainers using this term

• Paper Walkthrough: It Takes Two to Match — Co-Evolving Both Sides of Retrieval with RLIt Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning
• Paper Explained — GenFirst: Let Generation Shape the Latent Space First, Reconstruction SecondGenFirst: Generation Before Reconstruction for Stable End-to-End Latent Generative Modeling
• The Mathematics of Diffusion — Generation Seen Through Scores and SDEsGenerative Modeling by Estimating Gradients of the Data Distribution