SRVAU-R1: Enhancing Video Anomaly Understanding via Reflection-Aware Learning
On the map: LLM reasoning → Chain-of-thought reasoning for driving
Abstract (arXiv)
Multi-modal large language models (MLLMs) have demonstrated significant progress in reasoning capabilities and shown promising effectiveness in video anomaly understanding (VAU) tasks. However, existing MLLM-based approaches remain largely focused on surface-level descriptions of anomalies, lacking deep reasoning over abnormal behaviors like explicit self-reflection and self-correction. To address that, we propose Self-Reflection-Enhanced Reasoning for Video Anomaly Understanding (SRVAU-R1), a reflection-aware learning framework that incorporates reflection in MLLM reasoning. Specifically, SRVAU-R1 introduces the first reflection-oriented Chain-of-Thought dataset tailored for VAU, providing structured supervision with initial reasoning, self-reflection, and revised reasoning. Based on that, it includes a novel reflection-aware learning paradigm with supervised fine-tuning and reinforcement fine-tuning to enhance multi-modal reasoning for VAU. Extensive experiments on multiple video anomaly benchmarks demonstrate that SRVAU-R1 consistently outperforms existing methods, achieving significant improvements in both temporal anomaly localization accuracy and reasoning quality.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Chain-of-thought reasoning for driving, one of 17 directions in LLM reasoning. Builds on 38 of its 49 references that are in the map. First posted on arXiv 1 Feb 2026.
Builds on
38 of 49 references.
Closest ancestors · 1
Same direction, newest first
Other references · 37
- 2025Qwen3 Technical Reportcited by 8,195
- 2025Qwen2.5-VL Technical Reportcited by 5,418
- 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,317
- 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scalecited by 2,678
- 2025Understanding R1-Zero-Like Training: A Critical Perspectivecited by 1,276
- 2025Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?cited by 948
- 2025Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Modelscited by 653
- 2018Real-world Anomaly Detection in Surveillance Videoscited by 570
- 2019Memorizing Normality to Detect Anomaly: Memory-augmented Deep Autoencoder for Unsupervised Anomaly Detectioncited by 560
- 2025Visual-RFT: Visual Reinforcement Fine-Tuningcited by 501
- 2025Video-R1: Reinforcing Video Reasoning in MLLMscited by 464
- 2024Training Language Models to Self-Correct via Reinforcement Learningcited by 405
25 more
- 2017Future Frame Prediction for Anomaly Detection -- A New Baselinecited by 391
- 2025Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRscited by 373
- 2025VAPO: Efficient and Reliable Reinforcement Learning for Advanced Reasoning Taskscited by 281
- 2024Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimizationcited by 238
- 2020Not only Look, but also Listen: Learning Multimodal Violence Detection under Weak Supervisioncited by 235
- 2021Weakly-supervised Video Anomaly Detection with Robust Temporal Feature Magnitude Learningcited by 226
- 2025R1-Zero's "Aha Moment" in Visual Reasoning on a 2B Non-SFT Modelcited by 163
- 2021A Hybrid Video Anomaly Detection Framework via Memory-Augmented Flow Reconstruction and Flow-Guided Frame Predictioncited by 129
- 2023Dual Memory Units with Uncertainty Regulation for Weakly Supervised Video Anomaly Detectioncited by 88
- 2022MGFN: Magnitude-Contrastive Glance-and-Focus Network for Weakly-Supervised Video Anomaly Detectioncited by 87
- 2024Harnessing Large Language Models for Training-free Video Anomaly Detectioncited by 85
- 2022CLIP-TSA: CLIP-Assisted Temporal Self-Attention for Weakly-Supervised Video Anomaly Detectioncited by 67
- 2024Holmes-VAU: Towards Long-term Video Anomaly Understanding at Any Granularitycited by 66
- 2024Holmes-VAD: Towards Unbiased and Explainable Video Anomaly Detection via Multi-modal LLMcited by 63
- 2020Few-shot Scene-adaptive Anomaly Detectioncited by 62
- 2025SRPO: Enhancing Multimodal LLM Reasoning via Reflection-Aware Reinforcement Learningcited by 53
- 2024Advancing Video Anomaly Detection: A Concise Review and a New Datasetcited by 48
- 2024Hawk: Learning to Understand Open-World Video Anomaliescited by 47
- 2024Video Anomaly Detection and Explanation via Large Language Modelscited by 45
- 2025VAU-R1: Advancing Video Anomaly Understanding via Reinforcement Fine-Tuningcited by 24
- 2025Why Reasoning Matters? A Survey of Advancements in Multimodal Reasoning (v1)cited by 23
- 2024Learnable Expansion of Graph Operators for Multi-Modal Feature Fusioncited by 22
- 2025Plan Then Action:High-Level Planning Guidance Reinforcement Learning for LLM Reasoningcited by 17
- 2024Beyond Euclidean: Dual-Space Representation Learning for Weakly Supervised Video Violence Detectioncited by 10
- 2025Sherlock: Towards Multi-scene Video Abnormal Event Extraction and Localization via a Global-local Spatial-sensitive LLMcited by 10
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026Advancing Adaptive Multi-Stage Video Anomaly Reasoning: A Benchmark Dataset and Method
- 2026PrismVAU: Prompt-Refined Inference System for Multimodal Video Anomaly Understanding
- 2026Reasoning-Guided Grounding: Elevating Video Anomaly Detection through Multimodal Large Language Models
- 2026Reflect-R1: Evidence-Driven Reflection for Self-Correction in Long Video Understanding
- 2025VADER: Towards Causal Video Anomaly Understanding with Relation-Aware Large Language Models
- 2026Cog-VADU: A Training-Free Cognitive Reasoning Framework for Video Anomaly Detection and Understanding
- 2025Sparse Reasoning is Enough: Biological-Inspired Framework for Video Anomaly Detection with Large Pre-trained Models
- 2024Follow the Rules: Reasoning for Video Anomaly Detection with Large Language Models