#Model Distillation

Test-Time Distillation Nearly Doubles Model Performance

New research shows stronger AI models can guide weaker ones at inference time, nearly doubling performance without retraining through 'scaffolding'.