Research map / Continual learning / Online and replay-based
Look-ahead Meta Learning for Continual 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 |
|---|---|---|---|
| add learned per-parameter asynchronous rates for meta-updates (ablation) | C-MAML · CIFAR-100 Multiple-Pass | RA: 65.44 { $\pm$ 0.99} -> \textbf{70.08 { $\pm$ 0.66}} | Helped |
| add learned per-parameter asynchronous rates for meta-updates (ablation) | C-MAML · CIFAR-100 Single-Pass | RA: 55.57 { $\pm$ 0.94} -> \textbf{61.18 { $\pm$ 1.44}} | Helped |
| add learned per-parameter asynchronous rates for meta-updates (ablation) | C-MAML · TinyImagenet Multiple-Pass | RA: 61.93 { $\pm$ 1.55} -> \textbf{66.99 { $\pm$ 1.65}} | Helped |
| add learned per-parameter asynchronous rates for meta-updates (ablation) | C-MAML · TinyImagenet Single-Pass | RA: 48.77 { $\pm$ 1.26} -> 52.59 { $\pm$ 1.35} | Helped |
| make learned meta-update rates asynchronous rather than fixed (ablation) | Sync · CIFAR-100 Multiple-Pass | RA: 67.06 { $\pm$ 0.62} -> \textbf{70.08 { $\pm$ 0.66}} | Helped |
| make learned meta-update rates asynchronous rather than fixed (ablation) | Sync · CIFAR-100 Single-Pass | RA: 58.99 { $\pm$ 1.40} -> \textbf{61.18 { $\pm$ 1.44}} | Helped |
| make learned meta-update rates asynchronous rather than fixed (ablation) | Sync · TinyImagenet Multiple-Pass | RA: 65.40 { $\pm$ 1.40} -> \textbf{66.99 { $\pm$ 1.65}} | Helped |
| make learned meta-update rates asynchronous rather than fixed (ablation) | Sync · TinyImagenet Single-Pass | RA: \textbf{52.84 { $\pm$ 2.55}} -> 52.59 { $\pm$ 1.35} | No clear effect |
| add meta-learning to LR-modulated ER updates (ablation) | La-ER · CIFAR-100 Multiple-Pass | RA: 67.17 { $\pm$ 1.14} -> \textbf{70.08 { $\pm$ 0.66}} | Helped |
| add meta-learning to LR-modulated ER updates (ablation) | La-ER · TinyImagenet Multiple-Pass | RA: 54.76 { $\pm$ 1.94} -> \textbf{66.99 { $\pm$ 1.65}} | Helped |
| add meta-learning to LR-modulated ER updates (ablation) | La-ER · TinyImagenet Single-Pass | RA: 44.75 { $\pm$ 1.96} -> 52.59 { $\pm$ 1.35} | Helped |
| add asynchronous LR modulation (ablation) | C-MAML · MNIST Permutations | RA: \textbf{74.54 { $\pm$ 0.54}} -> 74.34 { $\pm$ 0.67} | No clear effect |
Closest papers by meaning
- Learning to Continually Learn with the Bayesian Principle (ICML 2024 · similarity 0.91)
- Improving Continual Learning Performance and Efficiency with Auxiliary Classifiers (ICML 2025 · similarity 0.88)
- LoRanPAC: Low-rank Random Features and Pre-trained Models for Bridging Theory and Practice in Continual Learning (ICLR 2025 · similarity 0.87)
- Trust Region Continual Learning as an Implicit Meta-Learner (NeurIPS 2026 · similarity 0.87)
- Re-evaluating Continual Learning with Few-Shot Adaptation (NeurIPS 2026 · similarity 0.87)
- Train with Perturbation, Infer after Merging: A Two-Stage Framework for Continual Learning (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.