Research map / Label-efficient and robust learning / Test-time adaptation
Fast Context Adaptation via Meta-Learning
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 |
|---|---|---|---|
| adapt task-specific context parameters instead of the whole network | CAVIA · Sine curve regression | MSE: 0.23 (± 0.02) -> 0.19 (± 0.02) | Helped |
| increase the number of context parameters from 1 to 4 (ablation) | CAVIA · Sine curve regression | MSE: 0.84 (± 0.06) -> 0.19 (± 0.02) | Helped |
| increase context parameters from 4 to 50 (ablation) | CAVIA · Sine curve regression | MSE: 0.19 (± 0.02) -> 0.19 (± 0.02) | No clear effect |
| use CAVIA instead of MAML | Mini-ImageNet · 5-way 1-shot | accuracy: 48.07 ± 1.75 % -> 51.82 ± 0.65% | Helped |
| use CAVIA instead of MAML | Mini-ImageNet · 5-way 5-shot | accuracy: 63.15 ± 0.91 % -> 65.85 ± 0.55% | Helped |
| increase CAVIA network width from 32 to 512 filters (ablation) | CAVIA · Mini-ImageNet, 5-way 1-shot | accuracy: 47.24 ± 0.65% -> 51.82 ± 0.65% | Helped |
| increase CAVIA network width from 32 to 512 filters (ablation) | CAVIA · Mini-ImageNet, 5-way 5-shot | accuracy: 59.05 ± 0.54 % -> 65.85 ± 0.55% | Helped |
| omit backpropagation through the context update (ablation) | CAVIA (512) · Mini-ImageNet, 5-way 1-shot | accuracy: 51.82 ± 0.65% -> 49.92 ± 0.68% | Hurt |
| omit backpropagation through the context update (ablation) | CAVIA (512) · Mini-ImageNet, 5-way 5-shot | accuracy: 65.85 ± 0.55% -> 63.59 ± 0.57% | Hurt |
| increase MAML network width from 32 to 64 filters (ablation) | MAML · Mini-ImageNet, 5-way 1-shot | accuracy: 48.07 ± 1.75 % -> 44.70 ± 1.69% | Hurt |
| increase MAML network width from 32 to 64 filters (ablation) | MAML · Mini-ImageNet, 5-way 5-shot | accuracy: 63.15 ± 0.91 % -> 61.87 ± 0.93% | Hurt |
| use CAVIA instead of MAML | CelebA image completion · random pixels, 100 points | pixel-wise MSE: 0.017 -> 0.014 | Helped |
Closest papers by meaning
- Towards Effective Context for Meta-Reinforcement Learning: an Approach based on Contrastive Learning (AAAI 2021 · similarity 0.88)
- Learning to Learn with Contrastive Meta-Objective (NeurIPS 2025 · similarity 0.88)
- Provable Meta-Learning with Low-Rank Adaptations (NeurIPS 2025 · similarity 0.87)
- MetaCARD: Meta-Reinforcement Learning with Task Uncertainty Feedback via Decoupled Context-Aware Reward and Dynamics Components (AAAI 2024 · similarity 0.87)
- AMAGO-2: Breaking the Multi-Task Barrier in Meta-Reinforcement Learning with Transformers (NeurIPS 2024 · similarity 0.86)
- Sparse Autoencoders Reveal Temporal Difference Learning in Large Language Models (ICLR 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.