Research map / Continual learning / Online and replay-based
Continual Learning via Bit-Level Information Preserving
Check 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 BLIP instead of UCB | Continual learning model · MNIST-5 | ACC (%): 99.63 -> 99.64 | No clear effect |
| Use BLIP instead of UCB | Continual learning model · PMNIST | ACC (%): 97.42 -> 97.31 | Hurt |
| Use BLIP instead of HAT | Continual learning model · Alternating CIFAR10/100 | ACC (%): 80.22 -> 74.70 | Hurt |
| Use BLIP instead of ACL-R | Continual learning model · 20-Split mini-ImageNet | ACC (%): 62.07 ± 0.51 -> 65.69 ± 0.87 | Helped |
| Use BLIP instead of HAT | Continual learning model · 20-Split mini-ImageNet | BWT (%): -0.04 ± 0.03 -> -1.05 ± 0.42 | Hurt |
| Use BLIP instead of ACL | Continual learning model · Sequence of 5 tasks | ACC (%): 78.55 ± 0.29 -> 82.87 ± 1.43 | Helped |
| Use BLIP instead of ACL | Continual learning model · Sequence of 5 tasks | BWT (%): -0.01 ± 0.15 -> -0.13 ± 0.324 | Hurt |
| Use ResNet-18 instead of AlexNet (ablation) | BLIP · 20-Split mini-ImageNet | ACC (%): 65.69 ± 0.87 -> 65.94 ± 1.36 | No clear effect |
| Set F0 to 1 × 10^-14 instead of 5 × 10^-16 (ablation) | BLIP · 20-Split mini-ImageNet | BWT (%): -1.05 ± 0.42 -> -3.60 ± 0.59 | Hurt |
| Set F0 to 1 × 10^-16 instead of 5 × 10^-16 (ablation) | BLIP · 20-Split mini-ImageNet | ACC (%): 65.69 ± 0.87 -> 64.78 ± 0.96 | Hurt |
| Set F0 to 5 × 10^-17 instead of 5 × 10^-16 (ablation) | BLIP · 20-Split mini-ImageNet | ACC (%): 65.69 ± 0.87 -> 64.29 ± 0.82 | Hurt |
Closest papers by meaning
- Do Your Best and Get Enough Rest for Continual Learning (CVPR 2025 · similarity 0.90)
- Model Sensitivity Aware Continual Learning (NeurIPS 2024 · similarity 0.90)
- STAR: Stability-Inducing Weight Perturbation for Continual Learning (ICLR 2025 · similarity 0.89)
- A Unified and General Framework for Continual Learning (ICLR 2024 · similarity 0.89)
- Train with Perturbation, Infer after Merging: A Two-Stage Framework for Continual Learning (NeurIPS 2025 · similarity 0.88)
- Learning to Continually Learn with the Bayesian Principle (ICML 2024 · 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.