Advancing Adaptive Multi-Stage Video Anomaly Reasoning: A Benchmark Dataset and Method

TL;DR

This work defines Video Anomaly Reasoning (VAR) to move beyond detection toward structured, multi-stage reasoning for anomalous events. It introduces Vad-Reasoning-Plus, a large-scale dataset with PerCoAct-CoT annotations and 37 reasoning templates, along with adaptive hybrid reasoning and an anomaly-aware RL framework (A2-GRPO) to train Vad-R1-Plus. The approach achieves superior performance in both reasoning quality and answer accuracy, demonstrating robust, risk-aware, and stage-consistent anomaly understanding. Overall, the paper provides a principled benchmark and learning framework that enables reasoning-centric video intelligence with practical safety implications.

Abstract

Recent progress in reasoning capabilities of Multimodal Large Language Models(MLLMs) has highlighted their potential for performing complex video understanding tasks. However, in the domain of Video Anomaly Detection and Understanding (VAD&U), existing MLLM-based methods are largely limited to anomaly localization or post-hoc description, lacking explicit reasoning processes, risk awareness, and decision-oriented interpretation. To address this gap, we define a new task termed Video Anomaly Reasoning (VAR), which elevates video anomaly analysis from descriptive understanding to structured, multi-stage reasoning. VAR explicitly requires models to perform progressive reasoning over anomalous events before answering anomaly-related questions, encompassing visual perception, causal interpretation, and risk-aware decision making. To support this task, we present a new dataset with 8,641 videos, where each video is annotated with diverse question types corresponding to different reasoning depths, totaling more than 50,000 samples, making it one of the largest datasets for video anomaly. The annotations are based on a structured Perception-Cognition-Action Chain-of-Thought (PerCoAct-CoT), which formalizes domain-specific reasoning priors for video anomaly understanding. This design enables systematic evaluation of multi-stage and adaptive anomaly reasoning. In addition, we propose Anomaly-Aware Group Relative Policy Optimization to further enhance reasoning reliability under weak supervision. Building upon the proposed task and dataset, we develop an end-to-end MLLM-based VAR model termed Vad-R1-Plus, which supports adaptive hierarchical reasoning and risk-aware decision making. Extensive experiments demonstrate that the proposed benchmark and method effectively advance the reasoning capabilities of MLLMs on VAR tasks, outperforming both open-source and proprietary baselines.

Figures (8)

• Figure 1: Task comparison. (a) Existing explainable VAD methods typically output an additional explanation after anomaly confidence. (b) Our preliminary method Vad-R1 could conduct deep reasoning about abnormal events. (c) This paper presents Vad-R1-Plus, which supports adaptive hierarchical reasoning and answer open questions.
• Figure 2: Illustration of the proposed structured Chain-of-Thought and the adaptive hybrid reasoning mechanism.
• Figure 3: Statistical Analysis of the Vad-Reasoning-Plus Dataset
• Figure 4: Illustration of the proposed A2-GRPO.
• Figure 5: Performance comparison on different numbers of input tokens.