Research map / Time series / Anomaly detection
Towards Multimodal Time Series Anomaly Detection with Semantic Alignment and Condensed Interaction
Shiyan Hu, Jianxin Jin, Yang Shu, Peng Chen, Bin Yang, Chenjuan Guo
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 smoothness loss L_SM (ablation) | MindTS · MDT | Aff-F / V-PR / V-ROC: 86.29 -> 89.19 Aff-F; 63.23 -> 65.44 V-PR | Improved |
| add smoothness loss L_SM (ablation) | MindTS · Energy | Aff-F / V-PR / V-ROC: 73.26 -> 74.37 Aff-F; 48.53 -> 50.36 V-PR | Improved |
| add smoothness loss L_SM (ablation) | MindTS · Environment | Aff-F / V-PR / V-ROC: 83.40 -> 85.29 Aff-F; 91.58 -> 93.78 V-ROC | Improved |
| use original condenser instead of one conditioned on unmasked time patches (ablation) | MindTS · MDT | Aff-F / V-PR / V-ROC: 85.59 -> 89.19 Aff-F; 62.72 -> 65.44 V-PR | Improved |
| use original condenser instead of one conditioned on unmasked time patches (ablation) | MindTS · Environment | Aff-F / V-PR / V-ROC: 82.46 -> 85.29 Aff-F; 93.89 vs 93.78 V-ROC | Mixed |
| swap text encoder from DeepSeek to GPT2 (ablation) | MindTS · MDT | Aff-F / V-PR / V-ROC: 89.19 -> 85.56 Aff-F | Hurt |
| swap text encoder from DeepSeek to BERT (ablation) | MindTS · MDT | Aff-F / V-PR / V-ROC: 89.19 -> 82.30 Aff-F; 65.44 -> 53.68 V-PR | Hurt |
| swap text encoder from DeepSeek to LLAMA (ablation) | MindTS · Energy | Aff-F / V-PR / V-ROC: 74.37 -> 75.64 Aff-F; 74.44 -> 74.58 V-ROC | Mixed |
| remove exogenous text representation (ablation) | MindTS · Energy and MDT (figure) | Aff-F / V-PR / V-ROC | Hurt |
| remove endogenous text representation (ablation) | MindTS · Energy and MDT (figure) | Aff-F / V-PR / V-ROC | Hurt |
| add time-text semantic alignment (ablation) | MindTS · six datasets (figure) | Aff-F / V-PR / V-ROC | Improved |
| add content condenser (ablation) | MindTS · six datasets (figure) | Aff-F / V-PR / V-ROC | Improved |
Closest papers by meaning
- Empowering Time Series Analysis with Large-Scale Multimodal Pretraining (NeurIPS 2026 · similarity 0.89)
- Supervision Recovery for Time Series Anomaly Detection via Counterfactual Pairing (NeurIPS 2026 · similarity 0.87)
- TRACE: Grounding Time Series in Context for Multimodal Embedding and Retrieval (NeurIPS 2025 · similarity 0.86)
- Time-MMD: Multi-Domain Multimodal Dataset for Time Series Analysis (NeurIPS 2024 · similarity 0.86)
- CrossAD: Time Series Anomaly Detection with Cross-scale Associations and Cross-window Modeling (NeurIPS 2025 · similarity 0.86)
- VerbalTS: Generating Time Series from Texts (ICML 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.
