MG-RWKV: Multi-Grained Context-Aware RWKV for Temporal Forgery Localization

2026-07-01

AI Summary

This work addresses the challenge that existing temporal forgery localization methods struggle to simultaneously model global authentic context and local forged discontinuities, often limited by the restricted receptive fields of CNNs or the high computational complexity of Transformers. To overcome this, we propose MG-RWKV, a multi-granularity efficient framework built upon RWKV that processes full-length sequences with linear O(T) complexity. The core innovations include bidirectional RWKV-based temporal modeling, multi-granularity mixture-of-experts with dynamic routing, cross-granularity consistency constraints, and a boundary-aware weighting mechanism, collectively enhancing both localization accuracy and interpretability. Extensive experiments on Lav-DF, TVIL, and Psynd benchmarks demonstrate state-of-the-art performance, significantly reducing computational overhead and false positive rates in authentic regions.

Abstract

Driven by Artificial Intelligence-Generated Content (AIGC), the authenticity of audio-visual content is facing severe challenges. Temporal Forgery Localization (TFL) aims to precisely identify manipulated segments within untrimmed sequences. However, existing methods are limited by CNNs' local receptive fields or Transformers' quadratic complexity, while emerging linear models often struggle to balance global authentic context compression with local abrupt forgery perception. To address this, we propose MG-RWKV, a multi-granularity framework that leverages the data-dependent state evolution of RWKV to achieve efficient full-sequence processing with O(T) complexity. Our framework features three core innovations: (1) a Bidirectional RWKV architecture that captures bidirectional temporal contexts without quadratic overhead; (2) a Multi-Granularity Mixture of Experts (MG-MoE) that performs dynamic routing over explicit temporal receptive fields, adaptively selecting granularities based on forgery duration to significantly enhance decision interpretability; and (3) Cross-Granularity Consistency (CGC), which aligns adjacent feature pyramid levels through hierarchical scale-wise pairing and spatial boundary-aware weighting, effectively reducing false positives in authentic regions. Extensive experiments on Lav-DF, TVIL, and Psynd datasets demonstrate that MG-RWKV achieves state-of-the-art performance with low computational cost.

Problem

• Temporal Forgery Localization
• AIGC
• RWKV
• Multi-Grained Context
• Audio-Visual Forgery

Innovation

• RWKV
• Temporal Forgery Localization
• Multi-Granularity Mixture of Experts
• Cross-Granularity Consistency
• Bidirectional Context Modeling