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

View on arXiv ← Back to list

Authors: Tianxiao Li, Zhenglin Huang, Haiquan Wen, Yiwei He, Shuchang Lyu, Baoyuan Wu, Guangliang Cheng

Published: 2025-08-06 15:08:16+00:00

AI Summary

RAIDX is a novel deepfake detection framework that integrates Retrieval-Augmented Generation (RAG) and Group Relative Policy Optimization (GRPO) to improve detection accuracy and provide interpretable explanations. It achieves state-of-the-art performance by incorporating external knowledge and generating fine-grained textual descriptions and saliency maps without manual annotations.

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.

Key findings

RAIDX achieves state-of-the-art deepfake detection accuracy on multiple benchmarks, outperforming existing methods. It demonstrates strong generalization to unseen generative models and robustness to common image distortions. The GRPO training significantly enhances the quality and granularity of the generated explanations.

Approach

RAIDX uses a Vision Transformer to extract image features. These features are used by a RAG module to retrieve similar images and their labels for contextual information. A partially trainable LLM, fine-tuned with GRPO, then generates textual explanations and saliency maps based on the combined image and text information.

Datasets

SID-Set (Real and Synthetic subsets), COCO, Flickr, and 18 additional datasets from the AntifakePrompt benchmark (various diffusion models, including SDXL, DALLE-3, DiffusionDB, and GLIDE). DFDC dataset mentioned.

Model(s)

Vision Transformer (ViT), Large Language Model (LLM) with LoRA adapters (Qwen2.5-VL mentioned), FAISS for efficient similarity search.

Author countries

United Kingdom, China

← Previous