Tools
DiScoFormer: One transformer for density and score, across distributions
DiScoFormer is a unified transformer architecture designed to handle both density estimation and score-based modeling across different probability distributions. The model enables a single framework t
DiScoFormer is a unified transformer architecture designed to handle both density estimation and score-based modeling across different probability distributions. The model enables a single framework to perform tasks typically requiring separate specialized approaches, potentially improving efficiency and generalization in generative modeling. This work from Allen Institute for AI represents an advance in flexible, distribution-agnostic deep learning architectures.
Source: Hugging Face | 2026-06-29