Research map / Label-efficient and robust learning / Knowledge distillation
Distilling Cross-Task Knowledge via Relationship Matching
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 |
|---|---|---|---|
| distill teacher triplet comparisons and local classifier knowledge | MobileNet student · CUB cross-task (100 classes) | accuracy: 70.83 -> 73.38 | Helped |
| use REFILLED rather than the strongest reported baseline | Wide-ResNet student · CIFAR-100 | accuracy: 76.62 -> 77.49 | Helped |
| use REFILLED rather than the strongest reported baseline | MobileNet student · CUB same-task | accuracy: 77.72 -> 78.95 | Helped |
| distill across different model families using REFILLED | MobileNet student with ResNet-110 teacher · CIFAR-100 | accuracy: 71.94 -> 73.81 | Helped |
| reuse a pretrained cross-task teacher through REFILLED | ConvNet student · MiniImageNet 10-shot 5-way | accuracy: 74.86 -> 76.42 | Helped |
| add triplet alignment to the student embedding (ablation) | REFILLED on Wide-ResNet · CIFAR-100 embedding | NMI: 51.77 -> 54.39 | Helped |
| use local rather than global knowledge distillation (ablation) | REFILLED on Wide-ResNet · CIFAR-100, student (40, 2) | accuracy: 77.08 -> 77.49 | Helped |
| add triplet-aligned embeddings and local classifier distillation (ablation) | REFILLED on CUB cross-task · CUB cross-task (100 classes) | accuracy | Helped |
Closest papers by meaning
- Knowledge Distillation from A Stronger Teacher (NeurIPS 2022 · similarity 0.89)
- Knowledge Distillation With the Reused Teacher Classifier (CVPR 2022 · similarity 0.88)
- Cross-Layer Distillation with Semantic Calibration (AAAI 2021 · similarity 0.88)
- Amalgamating Knowledge towards Comprehensive Classification (AAAI 2019 · similarity 0.87)
- Learning Task-Agnostic Representations through Multi-Teacher Distillation (NeurIPS 2025 · similarity 0.86)
- Cooperative Knowledge Distillation: A Learner Agnostic Approach (AAAI 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.