Research map / Label-efficient and robust learning / Knowledge distillation
Wasserstein Contrastive Representation Distillation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 global and local Wasserstein contrastive transfer | CIFAR-100 student networks · CIFAR-100 test, same and different teacher-student architectures | accuracy | Helped |
| Add KD loss to WCoRD | CIFAR-100 student networks · CIFAR-100 test, most architecture pairings | accuracy | Helped |
| Add KD loss to WCoRD | ResNet-50 → MobileNetV2 and ResNet-50 → VGG-8 · CIFAR-100 test | accuracy: 70.45 -> 70.12; 74.86 -> 74.68 | Hurt |
| Use global Wasserstein contrastive transfer alone (ablation) | WCoRD · CIFAR-100 test | accuracy | Helped |
| Use local Wasserstein contrastive transfer alone (ablation) | WCoRD · CIFAR-100 test | accuracy | Mixed |
| Remove local contrastive transfer (λ₂=0) (ablation) | WCoRD · CIFAR-100, ResNet-32x4 → ResNet-8x4 | accuracy: 75.45 -> 75.95 | Hurt |
| Set local-transfer weight λ₂ to 0.07 rather than 0 (ablation) | WCoRD · Cross-modal Tiny-ImageNet, ResNet-8x4 student | AUC: 83.43 -> 84.50 | Helped |
| Use WCoRD rather than CRD | ResNet-18 cross-modal student · Tiny-ImageNet linear probing, residual blocks 2–4 | accuracy: 63.64 -> 63.70; 73.76 -> 74.23; 74.75 -> 75.43 | Helped |
| Use WCoRD+KD rather than CRD | ResNet-18 student · ImageNet validation | Top-1 error: 28.83 -> 28.44 | Helped |
Closest papers by meaning
- Contrastive Representation Distillation (ICLR 2020 · similarity 0.93)
- Wasserstein Distance Rivals Kullback-Leibler Divergence for Knowledge Distillation (NeurIPS 2024 · similarity 0.90)
- WASD: Wasserstein-based Knowledge Distillation for Large Language Models (NeurIPS 2026 · similarity 0.89)
- Progressively Knowledge Distillation via Re-parameterizing Diffusion Reverse Process (AAAI 2024 · similarity 0.87)
- PLD: A Choice-Theoretic List-Wise Knowledge Distillation (NeurIPS 2025 · similarity 0.86)
- All You Need in Knowledge Distillation Is a Tailored Coordinate System (AAAI 2025 · 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.