DiScoFormer: One transformer for density and score, across distributions

The article introduces DiScoFormer, a unified transformer model capable of performing both density estimation and score-based generation tasks across various data distributions.

• The model utilizes a single architecture to handle multiple generative modeling objectives simultaneously.
• It demonstrates effectiveness across different types of data distributions without requiring separate specialized models.
• The approach aims to streamline the generative modeling pipeline by consolidating distinct tasks into one framework.