Research map / Label-efficient and robust learning / Knowledge distillation
Customizing Student Networks From Heterogeneous Teachers via Adaptive Knowledge Amalgamation
Chengchao Shen, Mengqi Xue, Xinchao Wang, Jie Song, Li Sun, Mingli Song
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 transfer bridge (ablation) | amalgamation student · CelebA black hair | accuracy (%): 87.4 -> 87.8 | Improved |
| add transfer bridge (ablation) | amalgamation student · CelebA mouth slightly open | accuracy (%): 91.3 -> 91.4 | Improved |
| add transfer bridge (ablation) | amalgamation student · CelebA brown hair | accuracy (%): 87.7 -> 88.0 | Improved |
| add teacher-selective learning (ablation) | amalgamation student · CelebA black hair | accuracy (%): 87.4 -> 87.8 | Improved |
| add teacher-selective learning (ablation) | amalgamation student · CelebA mouth slightly open | accuracy (%): 91.0 -> 91.4 | Improved |
| add teacher-selective learning (ablation) | amalgamation student · CelebA brown hair | accuracy (%): 87.8 -> 88.0 | Improved |
| use full method instead of plain knowledge distillation (ablation) | knowledge distillation baseline · CelebA mouth slightly open | accuracy (%): 90.2 -> 91.4 | Improved |
| use two-stage instead of one-shot amalgamation (ablation) | one-shot amalgamation · CelebA Black Hair | accuracy (%): 85.6 -> 87.6 | Improved |
| use two-stage instead of one-shot amalgamation (ablation) | one-shot amalgamation · CelebA Blond Hair | accuracy (%): 86.1 -> 95.1 | Improved |
| amalgamate component nets into a target net | component net · CelebA Smiling | accuracy (%): 90.5 -> 90.8 | Improved |
| amalgamate component nets into a target net | component net · Cars C1 | accuracy (%): 72.1 -> 72.5 | Improved |
| amalgamate more source nets into component net | component net · CelebA face attributes | accuracy (%) | Improved |
Closest papers by meaning
- Collaboration by Competition: Self-coordinated Knowledge Amalgamation for Multi-talent Student Learning (ECCV 2020 · similarity 0.90)
- Amalgamating Multi-Task Models with Heterogeneous Architectures (AAAI 2024 · similarity 0.88)
- Toward Student-oriented Teacher Network Training for Knowledge Distillation (ICLR 2024 · similarity 0.87)
- Less or More From Teacher: Exploiting Trilateral Geometry For Knowledge Distillation (ICLR 2024 · similarity 0.86)
- Hybrid Data-Free Knowledge Distillation (AAAI 2025 · similarity 0.86)
- Single-Teacher View Augmentation: Boosting Knowledge Distillation via Angular Diversity (NeurIPS 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.
