mlsanityData to 14 Oct 2026

Flashback: Memory-Driven Zero-shot, Real-time Video Anomaly Detection

arXiv 2505.15205 ↗First posted 21 May 2025cs.CVcited by 12

Not on the map. Why?

Abstract (arXiv)

Video Anomaly Detection (VAD) automatically identifies anomalous events from video, mitigating the need for human operators in large-scale surveillance deployments. However, two fundamental obstacles hinder real-world adoption: domain dependency and real-time constraints -- requiring near-instantaneous processing of incoming video. To this end, we propose Flashback, a zero-shot and real-time video anomaly detection paradigm. Inspired by the human cognitive mechanism of instantly judging anomalies and reasoning in current scenes based on past experience, Flashback operates in two stages: Recall and Respond. In the offline recall stage, an off-the-shelf LLM builds a pseudo-scene memory of both normal and anomalous captions without any reliance on real anomaly data. In the online respond stage, incoming video segments are embedded and matched against this memory via similarity search. By eliminating all LLM calls at inference time, Flashback delivers real-time VAD even on a consumer-grade GPU. On two large datasets from real-world surveillance scenarios, UCF-Crime and XD-Violence, we achieve 87.3 AUC (+7.0 pp) and 75.1 AP (+13.1 pp), respectively, outperforming prior zero-shot VAD methods by large margins.

Key facts. Cited by 12 papers in the map (data through 14 Oct 2026). Builds on 42 of its 61 references that are in the map. First posted on arXiv 21 May 2025.

Builds on

42 of 61 references.

References · 42

  1. 2020Language Models are Few-Shot Learnerscited by 32,419
  2. 2022Training language models to follow instructions with human feedbackcited by 15,971
  3. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  4. 2023Visual Instruction Tuningcited by 7,869
  5. 2023BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Modelscited by 6,137
  6. 2023Improved Baselines with Visual Instruction Tuningcited by 4,418
  7. 2022Flamingo: a Visual Language Model for Few-Shot Learningcited by 4,386
  8. 2021Finetuned Language Models Are Zero-Shot Learnerscited by 3,780
  9. 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,955
  10. 2023InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuningcited by 2,876
  11. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,724
  12. 2016Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answeringcited by 2,516
30 more
  1. 2015VQA: Visual Question Answeringcited by 2,506
  2. 2018CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledgecited by 1,822
  3. 2023ImageBind: One Embedding Space To Bind Them Allcited by 1,175
  4. 2019OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledgecited by 1,121
  5. 2023Kosmos-2: Grounding Multimodal Large Language Models to the Worldcited by 1,089
  6. 2016The LAMBADA dataset: Word prediction requiring a broad discourse contextcited by 992
  7. 2021Towards Total Recall in Industrial Anomaly Detectioncited by 776
  8. 2018Real-world Anomaly Detection in Surveillance Videoscited by 577
  9. 2025Perception Encoder: The best visual embeddings are not at the output of the networkcited by 326
  10. 2023AnomalyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detectioncited by 271
  11. 2016Learning Temporal Regularity in Video Sequencescited by 254
  12. 2020Not only Look, but also Listen: Learning Multimodal Violence Detection under Weak Supervisioncited by 238
  13. 2021Weakly-supervised Video Anomaly Detection with Robust Temporal Feature Magnitude Learningcited by 230
  14. 2023VadCLIP: Adapting Vision-Language Models for Weakly Supervised Video Anomaly Detectioncited by 123
  15. 2022Generative Cooperative Learning for Unsupervised Video Anomaly Detectioncited by 102
  16. 2022MGFN: Magnitude-Contrastive Glance-and-Focus Network for Weakly-Supervised Video Anomaly Detectioncited by 89
  17. 2022Anomaly Detection in Autonomous Driving: A Surveycited by 87
  18. 2024Harnessing Large Language Models for Training-free Video Anomaly Detectioncited by 87
  19. 2022CLIP-TSA: CLIP-Assisted Temporal Self-Attention for Weakly-Supervised Video Anomaly Detectioncited by 67
  20. 2019GODS: Generalized One-class Discriminative Subspaces for Anomaly Detectioncited by 64
  21. 2024Holmes-VAD: Towards Unbiased and Explainable Video Anomaly Detection via Multi-modal LLMcited by 64
  22. 2023Open-Vocabulary Video Anomaly Detectioncited by 57
  23. 2025EventVAD: Training-Free Event-Aware Video Anomaly Detectioncited by 36
  24. 2022DyAnNet: A Scene Dynamicity Guided Self-Trained Video Anomaly Detection Networkcited by 26
  25. 2024Multi-modal Preference Alignment Remedies Degradation of Visual Instruction Tuning on Language Modelscited by 26
  26. 2023Exploring Diffusion Models for Unsupervised Video Anomaly Detectioncited by 23
  27. 2023Unsupervised Video Anomaly Detection with Diffusion Models Conditioned on Compact Motion Representationscited by 19
  28. 2020Video Anomaly Detection for Smart Surveillancecited by 16
  29. 2025AnyAnomaly: Zero-Shot Customizable Video Anomaly Detection with LVLMcited by 15
  30. 2025AVadCLIP: Audio-Visual Collaboration for Robust Video Anomaly Detectioncited by 7

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 2; 2026: 10.

12 papers.

  1. 2025Unlocking Vision-Language Models for Video Anomaly Detection via Fine-Grained Promptingcited by 8
  2. 2026Em-Garde: A Propose-Match Framework for Proactive Streaming Video Understandingcited by 7
  3. 2026Reasoning-Guided Grounding: Elevating Video Anomaly Detection through Multimodal Large Language Modelscited by 2
  4. 2026The Unreasonable Effectiveness of VLMs for Zero-shot Procedural Mistake Detectioncited by 2
  5. 2025TRACES: Temporal Recall with Contextual Embeddings for Real-Time Video Anomaly Detectioncited by 1
  6. 2026Probe-VAD: Ordinal Likelihood Probing for Training-Free Video Anomaly Detectioncited by 1
  7. 2026Decision Readouts for Text-Mediated Video Anomaly Detection: An Exploratory Evaluation of Jev and Qwencited by 1
  8. 2026Geometry-Aware Semantic Reasoning for Training Free Video Anomaly Detectioncited by 0
  9. 2026LATERN: Test-Time Context-Aware Explainable Video Anomaly Detectioncited by 0
  10. 2026NOVA: Normal-Side Modeling for Training-Free Zero-Shot Video Anomaly Detectioncited by 0
  11. 2026ReactVAU: A Slow-Fast Decoupled Framework for Streaming Video Anomaly Understandingcited by 0
  12. 2026PARSEE-VAD: Efficient Training-Free Online Video Anomaly Detection via Proposition-Aware Reasoning and Streaming Evidence Escalationcited by 0