GradAlign for Training-free Model Performance Inference

Re-assign community