Research map / Video / Video-language understanding
Streaming Video Crime Anticipation with Spatio-Temporal Causal Reasoning
Yusong Wang, Zheyuan Gu, Keyu Mao, Minghao Shao, Mingkun Xu, Prayag Tiwari, Jiawei Shao, Qingsong 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 STCRC causal reasoning dataset for training (ablation) | STCH + Qwen2-VL-7B · UCF-Crime | WF1: 23.03 -> 40.67 | Improved |
| add STCH hypergraph module (ablation) | Qwen2-VL-7B · UCF-Crime | WF1: 33.88 -> 40.67 | Improved |
| add STCH hypergraph module (ablation) | Qwen2-VL-7B · UCF-Crime | TimeDiff: 58.73 -> 55.80 | Improved |
| add Group Evolution hyperedges (ablation) | STCH · UCF-Crime | WF1: 36.64 -> 40.67 | Improved |
| add Causal Cascade Effect hyperedges (ablation) | STCH · UCF-Crime | AUC-S: 0.675 -> 0.718 | Improved |
| add memory bank for long-stream context (ablation) | STCH + Qwen2-VL-7B · UCF-Crime | AUC-L: 0.601 -> 0.692 | Improved |
| add STCRC tasks 1 and 2 (ablation) | STCH + Qwen2-VL-7B · UCF-Crime | TimeDiff: 58.36 -> 55.80 | Improved |
| add STCRC task 5 entity-event causality (ablation) | STCH + Qwen2-VL-7B · UCF-Crime | WF1: 29.97 -> 40.67 | Improved |
| fine-tune streaming baseline on STCRC | Flash-VStream · UCF-Crime | WF1: 23.83 -> 38.07 | Improved |
| fine-tune streaming baseline on STCRC | VideoLLM-online · UCF-Crime | WF1: 18.89 -> 24.89 | Improved |
| use smaller Qwen2-VL-3B backbone instead of 7B (ablation) | STCH + STCRC · UCF-Crime | WF1: 40.67 -> 33.15 | Hurt |
| use LLaVA-1.5-7B backbone instead of Qwen2-VL-7B (ablation) | STCH + STCRC · UCF-Crime | WF1: 40.67 -> 38.27 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Flash-VStream — Flash-VStream: Efficient Real-Time Understanding for Long Video Streams (ICCV 2025) · better on 3 of 3 shared results
- Holmes-VAU — Holmes-VAU: Towards Long-term Video Anomaly Understanding at Any Granularity (CVPR 2025) · 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
- Spatial-Temporal Augmentation for Crime Prediction (Student Abstract) (AAAI 2024 · similarity 0.86)
- Hawkes-Enhanced Spatial-Temporal Hypergraph Contrastive Learning Based on Criminal Correlations (AAAI 2024 · similarity 0.85)
- MECD: Unlocking Multi-Event Causal Discovery in Video Reasoning (NeurIPS 2024 · similarity 0.84)
- A Unified Reasoning Framework for Holistic Zero-Shot Video Anomaly Analysis (NeurIPS 2025 · similarity 0.83)
- MoniTor: Exploiting Large Language Models with Instruction for Online Video Anomaly Detection (NeurIPS 2025 · similarity 0.82)
- Towards Surveillance Video-and-Language Understanding: New Dataset Baselines and Challenges (CVPR 2024 · similarity 0.82)
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.
