Research map / Continual learning / Online and replay-based
Helpful or Harmful: Inter-Task Association in Continual Learning
Hyundong Jin, Eunwoo Kim
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 model search over old tasks (ablation) | H2 · Split CIFAR-10 task-incremental T5 | accuracy (%): 96.9 -> 97.3 | Improved |
| add model search over old tasks (ablation) | H2 · Split CIFAR-10 class-incremental T5 | accuracy (%): 79.6 -> 80.3 | Improved |
| add Fisher sensitivity measure (ablation) | H2 · Split CIFAR-10 task-incremental T5 | accuracy (%): 96.9 -> 97.3 | Improved |
| add Fisher sensitivity measure (ablation) | H2 · Split CIFAR-10 class-incremental T5 | accuracy (%): 78.7 -> 80.3 | Improved |
| add search and sensitivity (vs using all old parameters) (ablation) | PackNet-style framework · Split CIFAR-10 class-incremental T5 | accuracy (%): 77.4 -> 80.3 | Improved |
| add search and sensitivity (vs using all old parameters) (ablation) | PackNet-style framework · Split CIFAR-10 task-incremental T5 | accuracy (%): 96.8 -> 97.3 | Improved |
| add contrastive loss to meta classifier training (ablation) | H2 · Split CIFAR-10 class-incremental T5 | accuracy (%): 77.1 -> 80.3 | Improved |
| replace helpful-parameter selection with random selection of old parameters (ablation) | H2 · Split CIFAR-10 class-incremental T5 | accuracy (%): 80.3 -> 68.7 | Hurt |
| replace helpful-parameter selection with random selection of old parameters (ablation) | H2 · Split CIFAR-10 task-incremental T5 | accuracy (%): 97.3 -> 96.3 | Hurt |
| add search and sensitivity, reducing parameters used per task (ablation) | H2 · Split CIFAR-10 parameter consumption | parameters at T5: 4.2x higher than ours for T5 without them | Improved |
| H2 vs PackNet (all old parameters) in task-incremental | H2 · ImageNet + fine-grained avg | accuracy (%): 80.2 -> 82.8 | Improved |
| H2 vs TAC in task-incremental | H2 · Split CIFAR-10 task-incremental | accuracy (%): 96.4 -> 97.3 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Piggyback — Piggyback: Adapting a Single Network to Multiple Tasks by Learning to Mask Weights (ECCV 2018) · better on 3 of 3 shared results
- iCaRL — iCaRL: Incremental Classifier and Representation Learning (CVPR 2017) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Continual Learning with Global Alignment (NeurIPS 2024 · similarity 0.89)
- Disentangling and mitigating the impact of task similarity for continual learning (NeurIPS 2024 · similarity 0.88)
- Train with Perturbation, Infer after Merging: A Two-Stage Framework for Continual Learning (NeurIPS 2025 · similarity 0.87)
- Theory on Mixture-of-Experts in Continual Learning (ICLR 2025 · similarity 0.87)
- The Joint Effect of Task Similarity and Overparameterization on Catastrophic Forgetting — An Analytical Model (ICLR 2024 · similarity 0.87)
- Learn more, but bother less: parameter efficient continual learning (NeurIPS 2024 · 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.
