Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Towards Enabling Meta-Learning from Target Models
Read the paperCheck 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 |
|---|---|---|---|
| train under S/T protocol with 10% target models instead of S/Q | ProtoNet · miniImageNet 5-way 1-shot | accuracy: 65.30 -> 68.03 | Helped |
| train under S/T protocol with 10% target models instead of S/Q | ProtoNet · tieredImageNet 5-way 5-shot | accuracy: 84.68 -> 86.91 | Helped |
| train under S/T protocol with 10% target models instead of S/Q | MAML · tieredImageNet 5-way 5-shot | accuracy: 67.26 -> 70.02 | Helped |
| train under S/T protocol with 10% target models instead of S/Q | MAML · miniImageNet 5-way 1-shot | accuracy: 58.84 -> 60.06 | Helped |
| use 5% instead of 10% target models (ablation) | ProtoNet S/T · miniImageNet 5-way 1-shot | accuracy: 68.03 -> 67.35 | Hurt |
| select tasks by hardness and fine-tuned target models instead of random task selection (ablation) | ProtoNet S/T-10% · miniImageNet 5-way 1-shot | accuracy: 66.72 -> 68.03 | Helped |
| select tasks by hardness and fine-tuned target models instead of random task selection (ablation) | MAML S/T-10% · tieredImageNet 5-way 5-shot | accuracy: 68.63 -> 70.02 | Helped |
| fine-tune target models locally instead of using the pre-trained network for all tasks (ablation) | ProtoNet S/T-10% · tieredImageNet 5-way 5-shot | accuracy: 86.04 -> 86.91 | Helped |
| use pre-trained network as target model (non-optimal target) (ablation) | MAML S/T-10% · miniImageNet 5-way 5-shot | accuracy: 74.62 -> 75.88 | Helped |
| increase lambda in the S/T loss (ablation) | ProtoNet · miniImageNet 5-way | accuracy | Helped |
| apply S/T protocol with 10% target models to FEAT | FEAT · miniImageNet 5-way 1-shot | accuracy: 66.78 -> 68.23 | Helped |
| apply S/T protocol with 5% target models to FEAT | FEAT · miniImageNet 5-way 5-shot | accuracy: 82.05 -> 81.60 | Hurt |
Closest papers by meaning
- Submodular Meta-Learning (NeurIPS 2020 · similarity 0.88)
- Meta-Learning Requires Meta-Augmentation (NeurIPS 2020 · similarity 0.88)
- Large-Scale Meta-Learning with Continual Trajectory Shifting (ICML 2021 · similarity 0.88)
- Meta-Learning with Self-Improving Momentum Target (NeurIPS 2022 · similarity 0.88)
- Memory-Reduced Meta-Learning with Guaranteed Convergence (AAAI 2025 · similarity 0.88)
- Provable Meta-Learning with Low-Rank Adaptations (NeurIPS 2025 · similarity 0.87)
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.