Research map / Video / Video segmentation and tracking
ReactVAU: A Slow-Fast Decoupled Framework for Streaming Video Anomaly Understanding
Chia-Hui Chen, Shih-Ying Yeh, Fu-En Yang, Min-Hung Chen, Shang-Hong Lai
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 |
|---|---|---|---|
| fold frames into a 2x2 grid (SGF) without fine-tuning (ablation) | PaliGemma2-3B · XD-Violence | AP: 53.71 -> 43.42 | Hurt |
| fine-tune SGF on the Grid Image Dataset (ablation) | PaliGemma2-3B · XD-Violence | AP: 43.42 -> 79.55 | Improved |
| add Slow-Fast decoupling (only fast module tuned) (ablation) | StreamForest-7B fine-tuned · XD-Violence | AP: 75.92 -> 80.59 | Improved |
| add Slow-Fast decoupling (only fast module tuned) (ablation) | StreamForest-7B fine-tuned · XD-Violence | AUC: 92.82 -> 91.34 | Hurt |
| tune both fast and slow modules in decoupled framework (ablation) | Slow-Fast Decoupling · XD-Violence | AP: 80.59 -> 86.90 | Improved |
| add Anomaly-Aware Persistent Memory (AAPM) (ablation) | Slow-Fast Decoupling (Both Tuned) · XD-Violence | AP: 86.90 -> 88.50 | Improved |
| add Anomaly-Aware Persistent Memory (AAPM) (ablation) | Slow-Fast Decoupling (Both Tuned) · UCF-Crime | AUC: 87.17 -> 88.44 | Improved |
| add Anomaly-Aware Persistent Memory (AAPM) (ablation) | Slow-Fast Decoupling (Both Tuned) · XD-Violence | AUC: 94.16 -> 95.25 | Improved |
| fine-tune StreamForest on HIVAU/anomaly data (ablation) | StreamForest-7B · XD-Violence | AP: 49.24 -> 75.92 | Improved |
| fuse fast and slow scores (0.4/0.6) instead of replacing fast score with reasoning score (ablation) | ReactVAU · UCF-Crime | AUC: 87.28 -> 88.44 | Improved |
| add Fast Module gating of the 7B Slow Module | StreamForest · UCF-Crime | 7B LLM queries: 34,670 -> 15,955 | Improved |
| add Fast Module gating of the 7B Slow Module | StreamForest · UCF-Crime | weighted latency (ms/query): 216.1 -> 98.3 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MoniTor — MoniTor: Exploiting Large Language Models with Instruction for Online Video Anomaly Detection (NeurIPS 2025) · better on 3 of 3 shared results
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
- Holmes-VAU: Towards Long-term Video Anomaly Understanding at Any Granularity (CVPR 2025 · similarity 0.88)
- MoniTor: Exploiting Large Language Models with Instruction for Online Video Anomaly Detection (NeurIPS 2025 · similarity 0.87)
- A Unified Reasoning Framework for Holistic Zero-Shot Video Anomaly Analysis (NeurIPS 2025 · similarity 0.87)
- VADTree: Explainable Training-Free Video Anomaly Detection via Hierarchical Granularity-Aware Tree (NeurIPS 2025 · similarity 0.86)
- PANDA: Towards Generalist Video Anomaly Detection via Agentic AI Engineer (NeurIPS 2025 · similarity 0.85)
- Open-Vocabulary Video Anomaly Detection (CVPR 2024 · 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.
