Research map / Label-efficient and robust learning / OOD and anomaly detection
CL-Anomaly: Layer-Adaptive Mixture-of-Experts with Multimodal Large Language Model for Continual Learning in Anomaly Detection
Wen Dong, Zhao Wang, Shuangqing Zhang, Kai Sun, Ben Li, Guosen Xie, Caifeng Shan, Fang Zhao
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 PrivLoRA task-private experts (ablation) | CL-Anomaly · Anomaly-Instruct-125K continual (8 tasks) | Avg accuracy: 83.58 -> 86.34 | Improved |
| add PrivLoRA task-private experts (ablation) | CL-Anomaly · Anomaly-Instruct-125K continual (8 tasks) | Last accuracy: 81.60 -> 85.61 | Improved |
| add PrivLoRA task-private experts (ablation) | CL-Anomaly · Anomaly-Instruct-125K continual (8 tasks) | BWT: -3.70 -> -0.96 | Improved |
| add shared LoRA expert path (ablation) | CL-Anomaly · Anomaly-Instruct-125K continual (8 tasks) | Last accuracy: 82.63 -> 85.61 | Improved |
| add shared LoRA expert path (ablation) | CL-Anomaly · Anomaly-Instruct-125K continual (8 tasks) | BWT: -3.29 -> -0.96 | Improved |
| add layer-adaptive knowledge transfer (ablation) | CL-Anomaly · Anomaly-Instruct-125K continual (8 tasks) | Avg accuracy: 85.09 -> 86.34 | Improved |
| add layer-adaptive knowledge transfer (ablation) | CL-Anomaly · Anomaly-Instruct-125K continual (8 tasks) | Last accuracy: 84.06 -> 85.61 | Improved |
| add layer-adaptive knowledge transfer (ablation) | CL-Anomaly · Anomaly-Instruct-125K continual (8 tasks) | BWT: -1.38 -> -0.96 | Improved |
| CL-Anomaly instead of CL-MoE | CL-MoE · Anomaly-Instruct-125K continual (8 tasks) | Last accuracy: 78.08 -> 85.61 | Improved |
| CL-Anomaly instead of CL-MoE | CL-MoE · Anomaly-Instruct-125K continual (8 tasks) | Avg accuracy: 80.63 -> 86.34 | Improved |
| parameter isolation (HiDe-LLaVA) instead of sharing (CL-MoE) on medical BMAD | CL-MoE · BMAD (Last accuracy) | accuracy: 69.29 -> 87.88 | Improved |
| parameter sharing (CL-MoE) instead of isolation (HiDe-LLaVA) on class-incremental WebAD-1 | HiDe-LLaVA · WebAD-1 (Last accuracy) | accuracy: 75.31 -> 88.60 | Improved |
Closest papers by meaning
- Transfer-Based Semantic Anomaly Detection (ICML 2021 · similarity 0.88)
- Reasoning-Driven Anomaly Detection and Localization with Image-Level Supervision (CVPR 2026 · similarity 0.87)
- AnomalyMoE: Towards a Language-free Generalist Model for Unified Visual Anomaly Detection (AAAI 2026 · similarity 0.87)
- Unsupervised Continual Anomaly Detection with Contrastively-Learned Prompt (AAAI 2024 · similarity 0.86)
- VERA: Explainable Video Anomaly Detection via Verbalized Learning of Vision-Language Models (CVPR 2025 · similarity 0.86)
- One-for-More: Continual Diffusion Model for Anomaly Detection (CVPR 2025 · similarity 0.85)
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.
