Research map / Continual learning / Online and replay-based
Using Hindsight to Anchor Past Knowledge in 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 |
|---|---|---|---|
| Use HAL instead of MIR | Continual learning on Permuted MNIST · Permuted MNIST | Accuracy: 71.1 (\textpm 0.41) -> 73.6 (\textpm 0.31) | Helped |
| Use HAL instead of MIR | Continual learning on Permuted MNIST · Permuted MNIST | Forgetting: 0.11 (\textpm 0.01) -> 0.09 (\textpm 0.01) | Helped |
| Use HAL instead of MER | Continual learning on Rotated MNIST · Rotated MNIST | Accuracy: 66.0 (\textpm 2.04) -> 68.4 (\textpm 0.72) | Helped |
| Use HAL instead of MER | Continual learning on Rotated MNIST · Rotated MNIST | Forgetting: 0.23 (\textpm 0.01) -> 0.21 (\textpm 0.01) | Helped |
| Use HAL instead of MIR | Continual learning on Split CIFAR · Split CIFAR | Accuracy: 57.1 (\textpm 1.81) -> 60.4 (\textpm 0.54) | Helped |
| Use HAL instead of MIR | Continual learning on Split CIFAR · Split CIFAR | Forgetting: 0.12 (\textpm 0.01) -> 0.10 (\textpm 0.01) | Helped |
| Use HAL instead of MIR | Continual learning on Split miniImageNet · Split miniImageNet | Accuracy: 49.3 (\textpm 2.15) -> 51.6 (\textpm 2.02) | Helped |
| Use HAL instead of MIR | Continual learning on Split miniImageNet · Split miniImageNet | Forgetting: 0.12 (\textpm 0.01) -> 0.10 (\textpm 0.01) | Helped |
| Use hindsight anchors instead of oracle anchors | HAL on Permuted MNIST · Permuted MNIST | Accuracy: 73.9 (\textpm 0.41) -> 73.6 (\textpm 0.31) | No clear effect |
| Use hindsight anchors instead of oracle anchors | HAL on Split CIFAR · Split CIFAR | Accuracy: 61.1 (\textpm 0.94) -> 60.4 (\textpm 0.54) | No clear effect |
Closest papers by meaning
- Layerwise Proximal Replay: A Proximal Point Method for Online Continual Learning (ICML 2024 · similarity 0.90)
- Mitigating Catastrophic Forgetting in Online Continual Learning by Modeling Previous Task Interrelations via Pareto Optimization (ICML 2024 · similarity 0.89)
- Do Your Best and Get Enough Rest for Continual Learning (CVPR 2025 · similarity 0.88)
- Predicting the Susceptibility of Examples to Catastrophic Forgetting (ICML 2025 · similarity 0.88)
- A Unified and General Framework for Continual Learning (ICLR 2024 · similarity 0.88)
- Understanding the Forgetting of (Replay-based) Continual Learning via Feature Learning: Angle Matters (ICML 2025 · similarity 0.88)
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.