AgenticVAU: Multi-Agent Explore-Verify Reasoning for Video Anomaly Understanding

2026-08-04

AI Summary

This work addresses video anomaly understanding—a task requiring simultaneous anomaly detection, evidence localization, and causal explanation—by proposing a training-free multi-agent framework that overcomes the limitations of existing methods in generalization and evidence coverage due to reliance on task-specific training or restricted observation ranges. The approach models the problem as an explore-and-verify process, leveraging four specialized agents for rule construction, search planning, dense observation, and decision-making. A novel multi-agent collaboration mechanism, centered on an anchor registry serving as shared evidence memory, enables structured role assignment, coordinated evidence management, and iterative cross-temporal reasoning. Evaluated on the ECVA, UCF-Crime, and MSAD subsets of VAU-Bench, the method significantly outperforms zero-shot and reinforcement learning baselines, demonstrating its effectiveness.

Abstract

Video anomaly understanding (VAU) focuses on comprehensively interpreting abnormal events in videos, requiring models to identify anomalous occurrences, discover their supporting evidence, and explain the underlying causes beyond simple anomaly detection. Existing VAU methods often rely on specialized training or limited observations, restricting generalization or evidence coverage. Although single-agent alternatives support adaptive video observation, they still integrate exploration, observation, and decision-making within a unified reasoning process, offering limited role specialization and structured evidence coordination. To address these limitations, we present AgenticVAU, a training-free multi-agent framework that casts VAU as an explore--verify process, where the system first discovers potential anomalies and then verifies them through targeted observations. To achieve this, four specialized agents are introduced to handle visual-rule construction, search planning, video observation, and final decision, respectively. These agents communicate through an anchor registry, a shared evidence memory that binds each observation. Guided by this agent framework, AgenticVAU interleaves broad temporal exploration, dense local verification, and cross-interval comparison until sufficient evidence is collected. We conduct extensive experiments on the ECVA, UCF-Crime, and MSAD subsets of VAU-Bench, the results show that AgenticVAU outperforms zero-shot inference and reinforcement learning-based baselines, demonstrating the value of multi-agent collaboration for video anomaly understanding.

Problem

• Video Anomaly Understanding
• Multi-Agent Reasoning
• Evidence Coordination
• Generalization
• Anomaly Explanation

Innovation

• multi-agent
• explore-verify reasoning
• video anomaly understanding
• training-free framework
• evidence coordination