Research map / Label-efficient and robust learning / Knowledge distillation
Personalized Education: Blind Knowledge Distillation
Xiang Deng, Jian Zheng, Zhongfei Zhang
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 |
|---|---|---|---|
| use MixPatch instead of Mixup for distillation samples (ablation) | KD · CIFAR-100 (WRN-40-2 -> WRN-16-2) | test accuracy (%): 75.56 -> 75.94 | Improved |
| use MixPatch instead of Mixup for distillation samples (ablation) | KD · CIFAR-100 (VGG-13 -> VGG-8) | test accuracy (%): 74.58 -> 74.93 | Improved |
| add adaptive BKR search over MixPatch parameters (ablation) | KD+MixPatch · CIFAR-100 (WRN-40-2 -> WRN-16-2) | test accuracy (%): 75.94 -> 76.57 | Improved |
| add adaptive BKR search over MixPatch parameters (ablation) | KD+MixPatch · CIFAR-100 (VGG-13 -> VGG-8) | test accuracy (%): 74.93 -> 75.41 | Improved |
| add Personalized Education to KD | KD · CIFAR-100 (ResNet32x4 -> ShuffleNetV2) | test accuracy (%): 74.45 -> 79.83 | Improved |
| add Personalized Education to KD | KD · CIFAR-100 (ResNet32x4 -> ResNet8x4) | teacher-student accuracy gap: 6.30 -> 3.36 | Improved |
| add Personalized Education to KD | KD · Tiny ImageNet (VGG-13 -> VGG-8) | top-1 accuracy (%): 60.19 -> 62.40 | Improved |
| replace PE by GAN-generated distillation data (GANKD) | KD · CIFAR-100 (ResNet-110 -> ResNet-32) | test accuracy (%): 73.08 -> 72.87 | Hurt |
| add PE to AB distillation | AB · CIFAR-100 (WRN-40-2 -> WRN-16-2) | test accuracy (%): 68.89 -> 72.20 | Improved |
| add PE to CRD distillation | CRD · CIFAR-100 (WRN-40-2 -> WRN-16-2) | test accuracy (%): 76.04 -> 76.81 | Improved |
| distill on simulated full data distribution instead of training points | KD · CIFAR-100 (ResNet32x4 -> ShuffleNetV2) | test accuracy (%): 74.45 -> 81.64 | Improved |
| distill on simulated data distribution for incapable student SN2 | KD · CIFAR-100 (VGG-13 -> SN2) | test accuracy (%): 26.04 -> 25.58 | No clear effect |
Closest papers by meaning
- Can Students Beyond the Teacher? Distilling Knowledge from Teacher’s Bias (AAAI 2025 · similarity 0.90)
- Revisit the Essence of Distilling Knowledge through Calibration (ICML 2024 · similarity 0.88)
- PLD: A Choice-Theoretic List-Wise Knowledge Distillation (NeurIPS 2025 · similarity 0.88)
- Distilling What Matters: Confidence-Aware Selective Distillation for Large Language Models (NeurIPS 2026 · similarity 0.87)
- Toward Student-oriented Teacher Network Training for Knowledge Distillation (ICLR 2024 · similarity 0.87)
- Cooperative Knowledge Distillation: A Learner Agnostic Approach (AAAI 2024 · similarity 0.87)
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.
