A dynamic survey · v1.0
Language Models in
Video Anomaly Detection
A continuously maintained, agentic-AI-assisted survey covering methods, datasets, benchmarks, evaluation protocols, and emerging research directions, with human-reviewed releases.
182Papers
28Datasets
5Paradigms
v1.0Current Version
September 2026Last Updated
How to cite this resource
Cite the paper.
Reference the version.
Use the archival paper’s standard BibTeX citation. The maintained website version, URL, and access date are included as additional fields in that same entry.
Full citation guidance →Extended paper BibTeX
@article{mumcu2026dynamicvad,
title={Language Models in Video Anomaly Detection: A Dynamic Survey},
author={Mumcu, Furkan and Bekit, Lokman and Jones, Michael J. and Cherian, Anoop and Yilmaz, Yasin},
journal={IEEE Access},
year={2026},
version={v1.0},
url={https://dynamicvadsurvey.github.io/},
note={Website version v1.0, accessed __ACCESS_DATE__}
}Replace the clearly marked author, venue, and archival metadata once the final paper citation is supplied.
Current release · Version 1.0
Initial public release
62 structured methods28 datasets107 benchmark rows182 references
View release detailsTransparent maintenance
Assisted by agents.
Released by humans.
Agentic tools assist with discovery, summarization, routing, and localized updates. Public releases remain human-reviewed and versioned.
- 01Paper discovery
- 02Relevance screening
- 03Technical extraction
- 04Taxonomy routing
- 05Human review
- 06Versioned release