Research map / Label-efficient and robust learning / Knowledge distillation
Why Self-Distillation Helps and Hurts: Denoising vs. Signal Forgetting
Mingqi Wu, Archer Yi Yang, Qiang Sun
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add more self-distillation iterations on fresh covariates (overparameterized regime) | iterative self-distillation, linear regression · Power-law covariance simulation (Sigma_ii=1/i) | test MSE | Mixed |
| increase initial label noise to sigma = 0.8 versus 0.4 | self-distillation with ResNet-50 · CIFAR-10 | test error | Improved |
| increase training set size n | self-distillation with ResNet-50 · CIFAR-10 | test error | Improved |
| renew a larger fraction alpha of covariates each iteration | self-distillation, linear simulation · Appendix I simulation | test risk dynamics | Mixed |
| filter unlabeled samples aligned with the estimated signal subspace (smaller q) | self-distillation, linear simulation · Appendix I simulation | minimum test error | Improved |
| mix a fraction alpha_mix of real noisy labels into each iteration | self-distillation, linear simulation · Appendix I simulation | long-term test risk | Improved |
| apply self-distillation when initial teacher has little label noise | self-distillation with ResNet-50 · CIFAR-10 (low noise) | test error | Hurt |
Closest papers by meaning
- Understanding Self-Distillation in the Presence of Label Noise (ICML 2023 · similarity 0.91)
- Optimal Unconstrained Self-Distillation in Ridge Regression: Strict Improvements, Precise Asymptotics, and One-Shot Tuning (ICML 2026 · similarity 0.89)
- The Effect of Optimal Self-Distillation in Noisy Gaussian Mixture Model (NeurIPS 2025 · similarity 0.88)
- Self-Distillation Amplifies Regularization in Hilbert Space (NeurIPS 2020 · similarity 0.88)
- Self-cognitive Denoising in the Presence of Multiple Noisy Label Sources (ICML 2024 · similarity 0.88)
- Diffusion Tuning: Transferring Diffusion Models via Chain of Forgetting (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
