CEAVAD: Training-Free Video Anomaly Detection via Contrastive Event Adjudication

Forum topic · 小凯 · 2026-08-11

Summary

CEAVAD (Contrastive Event Adjudication for training-free Video Anomaly Detection) is a new approach from researchers Wenti Yin, Xiang Wang, and Huaxin Zhang, introduced in arXiv paper 2508.03800. Video anomaly detection (VAD) aims to identify and temporally localize abnormal events in videos. While supervised methods require large amounts of in-domain annotations, existing training-free methods exploit the semantic knowledge and reasoning abilities of pretrained models—but these capabilities do not directly define an anomaly decision criterion. CEAVAD shifts the unit of inference from isolated anomaly concepts to falsifiable event hypotheses. It first builds hazard-benign event contrasts using public safety knowledge, pairing each hazardous mechanism with a generic normal description and a mechanism-specific benign counterpart. It then determines whether a target interval better supports the hazardous explanation or its benign competitor, producing a correctable contrastive boundary proposal, and finally adjudicates among competing explanations to decide if the hazard hypothesis withstands the video evidence, supporting both temporal localization and evidence-based explanation. Experiments on three widely used VAD benchmarks show state-of-the-art performance under the training-free paradigm.

Paper Overview

Field: Computer Vision Authors: Wenti Yin, Xiang Wang, Huaxin Zhang Published: 2026-08-11 arXiv: 2508.03800

Introduction

Video anomaly detection (VAD) aims to identify and temporally localize abnormal events in videos. Supervised methods learn anomaly decision boundaries from target-domain annotations but require substantial in-domain data. Existing training-free methods leverage the rich semantic knowledge and reasoning capabilities of pretrained models to interpret visual content, yet these capabilities do not directly define an anomaly decision criterion: richer anomaly descriptions better capture hazard resemblance without resolving abnormality.

Proposed Method: CEAVAD

The authors propose Contrastive Event Adjudication for training-free Video Anomaly Detection (CEAVAD), which shifts the unit of inference from isolated anomaly concepts to falsifiable event hypotheses and establishes an inference-time explanatory boundary through the interaction between competing explanations and video evidence.

The method works in three stages:

1. Hazard-benign event contrast construction: Using public safety knowledge, each hazardous mechanism is paired with a generic normal description and a mechanism-specific benign counterpart. 2. Contrastive boundary proposal: For a target interval, CEAVAD determines whether the hazardous explanation or its benign competitor is better supported by the evidence, yielding a correctable contrastive boundary proposal. 3. Adjudication: CEAVAD adjudicates among competing explanations to decide whether the hazard hypothesis withstands scrutiny against the video evidence, supporting both temporally localized anomaly detection and evidence-based explanations.

Results

Experiments on three widely used VAD benchmarks demonstrate that CEAVAD achieves state-of-the-art performance under the training-free paradigm.