Research map / Continual learning / Continual fine-tuning of LLMs
Energy-Structured Low-Rank Adaptation for Continual Learning
Longhua Li, Lei Qi, Qi Tian, Xin Geng
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 classifier alignment (ablation) | E2-LoRA · Cars-196 | Last-Acc: 63.34 -> 75.82 | Improved |
| add self-distillation (ablation) | E2-LoRA · Cars-196 | Last-Acc: 72.15 -> 75.82 | Improved |
| add self-distillation without classifier alignment (ablation) | E2-LoRA · ImageNet-R | Last-Acc: 80.28 -> 80.28 | No clear effect |
| add classifier alignment (ablation) | E2-LoRA · ImageNet-R | Last-Acc: 80.28 -> 82.52 | Improved |
| use E2-LoRA instead of O-LoRA with KD and CA (ablation) | LoRA CL backbone · ImageNet-R | Last-Acc: 77.92 -> 82.77 | Improved |
| use E2-LoRA instead of InfLoRA with KD and CA (ablation) | LoRA CL backbone · Cars-196 | Last-Acc: 69.73 -> 75.82 | Improved |
| dynamic rank allocation instead of full rank (ablation) | E2-LoRA · ImageNet-R (50 tasks) | Last-Acc: 68.52 -> 78.58 | Improved |
| dynamic rank allocation instead of fixed rank dout/50 (ablation) | E2-LoRA · ImageNet-R (10 tasks) | Last-Acc: 77.71 -> 82.77 | Improved |
| dynamic rank allocation instead of fixed rank dout/5 (ablation) | E2-LoRA · ImageNet-R (10 tasks) | Last-Acc: 82.01 -> 82.77 | Improved |
| add PCA energy-structured transformation (ablation) | E2-LoRA · ImageNet-R | Last-Acc: 78.93 -> 82.77 | Improved |
| add PCA energy-structured transformation (ablation) | E2-LoRA · CIFAR-100 | Last-Acc: 87.04 -> 92.13 | Improved |
| increase proxy samples beyond 16 (ablation) | E2-LoRA · ImageNet-R and CIFAR-100 | Last-Acc | No clear effect |
Closest papers by meaning
- PaLoRA: Paced Low-Rank Adaptation for Continual Learning (NeurIPS 2026 · similarity 0.91)
- C-LoRA: Continual Low-Rank Adaptation for Pre-trained Visual Models (NeurIPS 2026 · similarity 0.90)
- Learn more, but bother less: parameter efficient continual learning (NeurIPS 2024 · similarity 0.88)
- LoRA Subtraction for Drift-Resistant Space in Exemplar-Free Continual Learning (CVPR 2025 · similarity 0.87)
- Continuous Subspace Optimization for Continual Learning (NeurIPS 2025 · similarity 0.87)
- SD-LoRA: Scalable Decoupled Low-Rank Adaptation for Class Incremental Learning (ICLR 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.
