RAIDX: A Retrieval-Augmented Generation and GRPO Reinforcement Learning Framework for Explainable Deepfake Detection

2025-08-06

AI Summary

Current deepfake detection methods suffer from two critical limitations: poor interpretability and heavy reliance on labor-intensive manual annotations—black-box classifiers lack transparent decision rationales, while large-model-based approaches yield coarse-grained explanations and incur high annotation costs. To address these issues, we propose the first interpretable detection framework integrating Retrieval-Augmented Generation (RAG) with Groupwise Relative Policy Optimization (GRPO). RAG enhances discriminative accuracy by incorporating external domain knowledge, whereas GRPO enables end-to-end generation of fine-grained textual explanations and pixel-level saliency maps without requiring extensive labeled data. Our method achieves state-of-the-art detection performance across multiple benchmarks and simultaneously produces trustworthy, verifiable multimodal attribution outputs—including natural-language justifications and visual saliency maps—thereby significantly improving system transparency, auditability, and practical deployability.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Transparent, Interpretable, Explainable ML
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

The rapid advancement of AI-generation models has enabled the creation of hyperrealistic imagery, posing ethical risks through widespread misinformation. Current deepfake detection methods, categorized as face specific detectors or general AI-generated detectors, lack transparency by framing detection as a classification task without explaining decisions. While several LLM-based approaches offer explainability, they suffer from coarse-grained analyses and dependency on labor-intensive annotations. This paper introduces RAIDX (Retrieval-Augmented Image Deepfake Detection and Explainability), a novel deepfake detection framework integrating Retrieval-Augmented Generation (RAG) and Group Relative Policy Optimization (GRPO) to enhance detection accuracy and decision explainability. Specifically, RAIDX leverages RAG to incorporate external knowledge for improved detection accuracy and employs GRPO to autonomously generate fine-grained textual explanations and saliency maps, eliminating the need for extensive manual annotations. Experiments on multiple benchmarks demonstrate RAIDX's effectiveness in identifying real or fake, and providing interpretable rationales in both textual descriptions and saliency maps, achieving state-of-the-art detection performance while advancing transparency in deepfake identification. RAIDX represents the first unified framework to synergize RAG and GRPO, addressing critical gaps in accuracy and explainability. Our code and models will be publicly available.

Problem

• Lack of transparency in current deepfake detection methods
• Coarse-grained analyses in LLM-based explainable approaches
• Dependency on labor-intensive annotations for explainability

Innovation

• RAG enhances detection with external knowledge
• GRPO generates fine-grained explanations autonomously
• Unified framework combines RAG and GRPO