Research map / Label-efficient and robust learning / Knowledge distillation
EA-KD: Entropy-based Adaptive Knowledge Distillation
Chi-Ping Su, Ching-Hsun Tseng, Bin Pu, Lei Zhao, Jiewen Yang, Zhuangzhuang Chen, Shin-Jye Lee
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 entropy-based reweighting w_EA to KD loss | KD · CIFAR-100 (ResNet32x4->ResNet8x4) | accuracy: 73.33 -> 75.46 | Improved |
| add teacher-entropy weight w_base (ablation) | KD · CIFAR-100 (ResNet32x4->ResNet8x4) | accuracy: 73.33 -> 75.14 | Improved |
| add interaction weight w_interact only (ablation) | KD · CIFAR-100 (ResNet32x4->ResNet8x4) | accuracy: 73.33 -> 74.76 | Improved |
| combine w_base and w_interact into w_EA (ablation) | KD · CIFAR-100 (ResNet32x4->ResNet8x4) | accuracy: 75.14 -> 75.46 | Improved |
| invert teacher-entropy weighting (low entropy gets high weight) (ablation) | KD · CIFAR-100 (ResNet32x4->ResNet8x4) | accuracy: 73.33 -> 72.73 | Hurt |
| invert student-entropy weighting (ablation) | KD · CIFAR-100 (ResNet32x4->ResNet8x4) | accuracy: 73.33 -> 68.90 | Hurt |
| set entropy temperature T' to 3 instead of 2 (ablation) | EA-KD · CIFAR-100 (ResNet32x4->ResNet8x4) | accuracy: 73.90 -> 75.46 | Improved |
| set entropy temperature T' to 3 instead of 4 (ablation) | EA-KD · CIFAR-100 (ResNet32x4->ResNet8x4) | accuracy: 75.22 -> 75.46 | Improved |
| add EA reweighting to MLD+LS | MLD+LS · CIFAR-100 (VGG13->MN-V2) | accuracy: 70.94 -> 70.67 | Hurt |
| add EA reweighting to FCFD | FCFD · CIFAR-100 (WRN-40-2->WRN-40-1) | accuracy: 75.46 -> 75.30 | Hurt |
| add EA reweighting to DKD | DKD · CIFAR-100 (beta sweep) | accuracy variance over beta: 0.31 -> 0.10 | Improved |
| add EA reweighting to KD with DeiT-B teacher | KD · Tiny-ImageNet (DeiT-B to ResNet8x4) | accuracy: 56.30 -> 58.36 | Improved |
Later papers that compared with this method
1 later accepted paper reports EA-KD as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Adaptive Conformal Guidance for Learning under Uncertainty (ICLR 2026 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PLD: A Choice-Theoretic List-Wise Knowledge Distillation (NeurIPS 2025 · similarity 0.90)
- DA-KD: Difficulty-Aware Knowledge Distillation for Efficient Large Language Models (ICML 2025 · similarity 0.89)
- ABKD: Pursuing a Proper Allocation of the Probability Mass in Knowledge Distillation via $\alpha$-$\beta$-Divergence (ICML 2025 · similarity 0.89)
- Can Students Beyond the Teacher? Distilling Knowledge from Teacher’s Bias (AAAI 2025 · similarity 0.88)
- TAID: Temporally Adaptive Interpolated Distillation for Efficient Knowledge Transfer in Language Models (ICLR 2025 · similarity 0.88)
- DeepKD: A Deeply Decoupled and Denoised Knowledge Distillation Trainer (NeurIPS 2025 · similarity 0.88)
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.
