PRPO: Paragraph-level Policy Optimization for Vision-Language Deepfake Detection
This addresses the challenge of reliable and interpretable deepfake detection for online safety and trust, representing a strong specific gain in a domain-specific area.
The paper tackled the problem of poor deepfake detection performance by multimodal large language models (LLMs) due to misaligned or hallucinatory reasoning, and introduced a reasoning-annotated dataset and a reinforcement learning algorithm (PRPO) that improved detection accuracy and achieved a reasoning score of 4.55/5.0.
The rapid rise of synthetic media has made deepfake detection a critical challenge for online safety and trust. Progress remains constrained by the scarcity of large, high-quality datasets. Although multimodal large language models (LLMs) exhibit strong reasoning capabilities, their performance on deepfake detection is poor, often producing explanations that are misaligned with visual evidence or hallucinatory. To address this limitation, we introduce a reasoning-annotated dataset for deepfake detection and propose Paragraph-level Relative Policy Optimization (PRPO), a reinforcement learning algorithm that aligns LLM reasoning with image content at the paragraph level. Experiments show that PRPO improves detection accuracy by a wide margin and achieves the highest reasoning score of 4.55/5.0. Ablation studies further demonstrate that PRPO significantly outperforms GRPO under test-time conditions. These results underscore the importance of grounding multimodal reasoning in visual evidence to enable more reliable and interpretable deepfake detection.