DiScoFormer: Transformer model for simultaneous density and score estimation across distributions
Research publication Provisional 95% confidence first seen
Researchers introduced DiScoFormer, a transformer-based model that estimates both probability density and score (gradient of log-density) from data in a single forward pass without requiring retraining. The model achieves 6.5x lower score error and 37x lower density error compared to kernel density estimation in 100 dimensions while generalizing across different distribution types.