DiScoFormer: One Transformer for Density and Score Across Distributions

A Hugging Face blog post introduces DiScoFormer, a framework that aims to use one Transformer to model both probability densities and score functions across distributions. The supplied metadata contains no abstract, benchmark results, model size, or implementation details, so claims about performance, scalability, supported distributions, and release status require verification in the original post.