Research map / Label-efficient and robust learning / Knowledge distillation
Black-Box Few-Shot Knowledge Distillation
Dang Nguyen, Sunil Gupta, Kien Do, Svetha Venkatesh
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 all three synthetic image types (mixup, CVAE-WD, CVAE-OOD) to standard KD (ablation) | FS-BBT · CIFAR-10 | accuracy: 58.96% -> 74.10% | Improved |
| add CVAE-OOD images to mixup + CVAE-WD (ablation) | FS-BBT · CIFAR-10 | accuracy: 72.60% -> 74.10% | Improved |
| add CVAE-WD images to mixup + CVAE-OOD (ablation) | FS-BBT · CIFAR-10 | accuracy: 73.25% -> 74.10% | Improved |
| add mixup images to CVAE-WD + CVAE-OOD (ablation) | FS-BBT · CIFAR-10 | accuracy: 70.63% -> 74.10% | Improved |
| add CVAE-WD images to mixup only (ablation) | FS-BBT · CIFAR-10 | accuracy: 71.67% -> 72.60% | Improved |
| add CVAE-OOD images to mixup only (ablation) | FS-BBT · CIFAR-10 | accuracy: 71.67% -> 73.25% | Improved |
| use mixup images only instead of standard KD on original images (ablation) | FS-BBT · CIFAR-10 | accuracy: 58.96% -> 71.67% | Improved |
| replace disqualified mixup images by CVAE images (FS-BBT vs BBKD) | FS-BBT · CIFAR-100 | accuracy: 53.41% -> 56.28% | Improved |
| replace BBKD's active-learning mixup pool with mixup + CVAE images | FS-BBT · Tiny-ImageNet | accuracy: 40.01% -> 43.29% | Improved |
| replace BBKD with FS-BBT on same LeNet5 architecture | FS-BBT* · Fashion-MNIST | accuracy: 80.90% -> 86.53% | Improved |
| replace BBKD with FS-BBT | FS-BBT · MNIST | accuracy: 98.74% -> 98.42% | No clear effect |
| use alpha threshold below 0.05 for filtering mixup images (ablation) | FS-BBT · CIFAR-10 | accuracy | Hurt |
Closest papers by meaning
- From Knowledge Distillation to Self-Knowledge Distillation: A Unified Approach with Normalized Loss and Customized Soft Labels (ICCV 2023 · similarity 0.91)
- Zero-Shot Knowledge Distillation from a Decision-Based Black-Box Model (ICML 2021 · similarity 0.90)
- Improving Knowledge Distillation via Regularizing Feature Direction and Norm (ECCV 2024 · similarity 0.90)
- Decoupled Knowledge Distillation (CVPR 2022 · similarity 0.89)
- Small Scale Data-Free Knowledge Distillation (CVPR 2024 · similarity 0.88)
- All You Need in Knowledge Distillation Is a Tailored Coordinate System (AAAI 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.
