Research map / Continual learning / Online and replay-based
C-Flat++: Towards a More Efficient and Powerful Framework for Continual Learning
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 C-Flat to the optimizer | Replay · CIFAR-100 B0_Inc10 | average accuracy: 58.09 -> 59.42 | Improved |
| add C-Flat to the optimizer | iCaRL · CIFAR-100 B0_Inc10 | average accuracy: 58.80 -> 59.75 | Improved |
| add C-Flat to the optimizer | WA · CIFAR-100 B0_Inc10 | average accuracy: 65.63 -> 67.00 | Improved |
| add C-Flat to the optimizer | PODNet · CIFAR-100 B0_Inc10 | average accuracy: 55.60 -> 57.00 | Improved |
| add C-Flat to the optimizer | DER · CIFAR-100 B0_Inc10 | average accuracy: 70.13 -> 71.35 | Improved |
| add C-Flat to the optimizer | FOSTER · CIFAR-100 B0_Inc10 | average accuracy: 65.84 -> 67.27 | Improved |
| add C-Flat to the optimizer | MEMO · CIFAR-100 B0_Inc10 | average accuracy: 68.49 -> 69.48 | Improved |
| add C-Flat to FOSTER on CUB | FOSTER · CUB B0_Inc20 | average accuracy: 31.84 -> 29.28 | Hurt |
| add C-Flat to WA on ImageNet-100 | WA · ImageNet-100 B50_Inc10 | average accuracy: 75.07 -> 74.03 | Hurt |
| C-Flat++ (adaptive 25.1% sharpness updates) instead of full C-Flat | C-Flat · Hybrid SGD/C-Flat study (avg accuracy) | average accuracy: 70.79 -> 70.39 | No clear effect |
| replace 25% of SGD steps with C-Flat (ablation) | SGD · Hybrid SGD/C-Flat study (avg accuracy) | average accuracy: 69.13 -> 69.91 | Improved |
| replace SGD flatness term FS with C-Flat | FS-DGPM · FS-DGPM series, average accuracy | average accuracy: 73.14 -> 73.72 | 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.
- FOSTER — FOSTER: Feature Boosting and Compression for Class-Incremental Learning (ECCV 2022) · better on 1 of 1 shared result
- iCaRL — iCaRL: Incremental Classifier and Representation Learning (CVPR 2017) · better on 1 of 1 shared result
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
- Make Continual Learning Stronger via C-Flat (NeurIPS 2024 · similarity 0.95)
- A Faster Path to Continual Learning (CVPR 2026 · similarity 0.92)
- Data Augmented Flatness-aware Gradient Projection for Continual Learning (ICCV 2023 · similarity 0.92)
- Model Sensitivity Aware Continual Learning (NeurIPS 2024 · similarity 0.87)
- A Unified and General Framework for Continual Learning (ICLR 2024 · similarity 0.86)
- LoRanPAC: Low-rank Random Features and Pre-trained Models for Bridging Theory and Practice in Continual Learning (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.
