Research map / Continual learning / Continual fine-tuning of LLMs
JANUS-LORA: A Balanced Low-Rank Adaptation for Continual Learning
Cheng Chen, Pengpeng Zeng, Yuyu Guo, Lianli Gao, Heng Tao Shen, Jingkuan Song
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 Online Estimation (OE) (ablation) | LoRA · ImageNet-R (T=10) | ACC: 64.24 -> 71.82 | Improved |
| add Online Estimation (OE) (ablation) | LoRA · ImageNet-R (T=10) | BWT: -21.85 -> -10.13 | Improved |
| add Gradient Rectification (GR) (ablation) | LoRA · ImageNet-R (T=10) | ACC: 64.24 -> 70.20 | Improved |
| add GR to OE (ablation) | Janus-LoRA w/o DML · ImageNet-R (T=10) | ACC: 71.82 -> 74.39 | Improved |
| add GR to OE (ablation) | Janus-LoRA w/o DML · ImageNet-R (T=10) | BWT: -10.13 -> -4.43 | Improved |
| add DML alone (ablation) | LoRA · ImageNet-R (T=10) | BWT: -21.85 -> -24.16 | Hurt |
| add DML alone (ablation) | LoRA · ImageNet-R (T=10) | ACC: 64.24 -> 64.47 | No clear effect |
| add DML to OE+GR (ablation) | Janus-LoRA · ImageNet-R (T=10) | ACC: 74.39 -> 75.78 | Improved |
| add DML to OE+GR (ablation) | Janus-LoRA · ImageNet-R (T=10) | BWT: -4.43 -> -6.05 | Hurt |
| add GR to OE+DML (ablation) | Janus-LoRA · ImageNet-R (T=10) | ACC: 73.61 -> 75.78 | Improved |
| increase OE rank k from 25 to 100 (ablation) | Janus-LoRA · ImageNet-R (T=10) | BWT: -7.41 -> -4.14 | Improved |
| plug GR into LoRA-GPM | LoRA-GPM · ImageNet-R (T=10) | ACC: +2.25% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- InfLoRA — InfLoRA: Interference-Free Low-Rank Adaptation for Continual Learning (CVPR 2024) · better on 7 of 7 shared results
- BiLoRA — BiLoRA: Almost-Orthogonal Parameter Spaces for Continual Learning (CVPR 2025) · better on 2 of 2 shared results
- Dual-Prompt — DualPrompt: Complementary Prompting for Rehearsal-Free Continual Learning (ECCV 2022) · 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
- C-LoRA: Continual Low-Rank Adaptation for Pre-trained Visual Models (NeurIPS 2026 · similarity 0.92)
- PaLoRA: Paced Low-Rank Adaptation for Continual Learning (NeurIPS 2026 · similarity 0.91)
- SD-LoRA: Scalable Decoupled Low-Rank Adaptation for Class Incremental Learning (ICLR 2025 · similarity 0.91)
- LoRA Subtraction for Drift-Resistant Space in Exemplar-Free Continual Learning (CVPR 2025 · similarity 0.90)
- CL-LoRA: Continual Low-Rank Adaptation for Rehearsal-Free Class-Incremental Learning (CVPR 2025 · similarity 0.89)
- SDS-LoRA: Overcoming Anisotropic Gradient Scaling in Low-Rank Adaptation (NeurIPS 2026 · similarity 0.89)
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.
