EVAS: Efficient Multimodal Temporal Forgery Localization via Audio-Visual Synergy and Steered Boundary Calibration

2026-07-05

AI Summary

This work addresses the challenge of localizing sparse, subtle, and cross-modally weakly correlated forged segments in long videos by proposing EVAS, an end-to-end multimodal framework. EVAS learns deep forensic representations through a multi-stage audio-visual collaboration mechanism, precisely refines forgery boundaries via a boundary-aware optimization strategy, and suppresses ambiguous regions using invalid-frame masking. To enhance generalization, the framework adopts a decoupled training paradigm and incorporates a lightweight HourglassFFN module to reduce computational overhead. Extensive experiments on three benchmark datasets demonstrate that EVAS achieves state-of-the-art performance in both average localization accuracy and recall, confirming its effectiveness for fine-grained temporal forgery localization.

Abstract

The rapid proliferation of artificial intelligence-generated content necessitates reliable multimodal forensics. Beyond video-level binary classification, precisely localizing sparsely distributed forged segments in long-form videos remains a critical challenge. This task is particularly difficult when manipulations are subtly embedded and cross-modal signals are weak and temporally diffuse. To address these challenges, we propose EVAS, an end-to-end multimodal framework for temporal forgery localization. At its core, a Multi-Stage Audio-Visual Synergy mechanism facilitates progressive cross-modal interaction to learn deep multimodal forensic representations and capture high-order semantic traces of sparse manipulations. Furthermore, we introduce a Boundary-Aware Refinement strategy to achieve steered boundary calibration. By incorporating invalid-frame masking, this strategy suppresses ambiguous regions and sharpens transition predictions. We adopt a decoupled training paradigm with auxiliary heads to disentangle representation learning from inference objectives, enhancing model generalization and stability. Additionally, a lightweight HourglassFFN is incorporated to reduce computational overhead. Extensive experiments demonstrate that EVAS achieves state-of-the-art average localization accuracy and average recall across three benchmark datasets, validating its effectiveness for fine-grained temporal forgery localization.

Problem

• temporal forgery localization
• multimodal forensics
• audio-visual synergy
• boundary calibration
• AI-generated content

Innovation

• Audio-Visual Synergy
• Boundary Calibration
• Temporal Forgery Localization
• Multimodal Forensics
• Decoupled Training