Authors
Loading...
Self-Competitive Distillation enhances test accuracy in lightweight models, suggesting effective training dynamics and improved performance.
Zhang et al. (2026) studied this question.