PRPO: Paragraph-level Policy Optimization for Vision-Language Deepfake Detection

• URL: http://arxiv.org/abs/2509.26272v2
• Date: Wed, 01 Oct 2025 12:27:21 GMT
• Title: PRPO: Paragraph-level Policy Optimization for Vision-Language Deepfake Detection
• Authors: Tuan Nguyen, Naseem Khan, Khang Tran, NhatHai Phan, Issa Khalil,
• Abstract summary: We introduce a reasoning-annotated dataset for deepfake detection and propose a reinforcement learning algorithm for deepfake detection.<n>Experiments show that PRPO improves detection accuracy by a wide margin and achieves the highest reasoning score of 4.55/5.0.<n>These results underscore the importance of grounding multimodal reasoning in visual evidence to enable more reliable and interpretable deepfake detection.
• Score: 7.944678181646657
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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.

Related papers

• MARE: Multimodal Alignment and Reinforcement for Explainable Deepfake Detection via Vision-Language Models [33.23839828166033]
  We propose multimodal alignment and reinforcement for explainable Deepfake detection via vision-language models, termed MARE.<n>MarE incorporates reinforcement learning from human feedback to incentivize the generation of reasoning content that adheres to human preferences.<n>MarE achieves state-of-the-art performance in terms of accuracy and reliability.
  arXiv  Detail & Related papers  (2026-01-28T09:44:31Z)
• AdversaRiskQA: An Adversarial Factuality Benchmark for High-Risk Domains [3.721111684544962]
  Hallucination in large language models (LLMs) contributes to spread of misinformation and diminished public trust.<n>We introduce AdversaRiskQA, the first verified and reliable benchmark systematically evaluating adversarial factuality.<n>We evaluate six open- and closed-source LLMs from the Qwen, GPT-OSS, and GPT families, measuring misinformation detection rates.
  arXiv  Detail & Related papers  (2026-01-21T22:47:59Z)
• VULPO: Context-Aware Vulnerability Detection via On-Policy LLM Optimization [2.6678231901651723]
  This paper introduces Vulnerability-Adaptive Policy Optimization (VULPO), an on-policy LLM reinforcement learning framework for context-aware vulnerability detection.<n>To support training and evaluation, we first construct ContextVul, a new dataset that augments high-quality function-level samples with lightweight method to extract repository-level context information.<n>To address the asymmetric difficulty of different vulnerability cases and mitigate reward hacking, VULPO incorporates label-level and sample-level difficulty-adaptive reward scaling.
  arXiv  Detail & Related papers  (2025-11-14T21:57:48Z)
• Repurposing Synthetic Data for Fine-grained Search Agent Supervision [81.95597592711688]
  LLM-based search agents are increasingly trained on entity-centric synthetic data.<n> prevailing training methods discard this rich entity information, relying instead on sparse, outcome-based rewards.<n>We introduce Entity-aware Group Relative Policy Optimization (E-GRPO), a novel framework that formulates a dense entity-aware reward function.
  arXiv  Detail & Related papers  (2025-10-28T17:50:40Z)
• Perception-Consistency Multimodal Large Language Models Reasoning via Caption-Regularized Policy Optimization [72.30168853571216]
  multimodal large language models excel at tasks that integrate visual perception with symbolic reasoning.<n>CapPO integrates two key mechanisms: (1) a caption-based consistency regularization, which minimizes the divergence between responses conditioned on raw images and those conditioned on captions, and (2) a KL-weighted advantage estimation scheme, which adaptively scales reinforcement signals to strengthen perceptually consistent trajectories.
  arXiv  Detail & Related papers  (2025-09-26T04:32:26Z)
• ThinkFake: Reasoning in Multimodal Large Language Models for AI-Generated Image Detection [51.93101033997245]
  Increasing realism of AI-generated images has raised serious concerns about misinformation and privacy violations.<n>We propose ThinkFake, a novel reasoning-based and generalizable framework for AI-generated image detection.<n>We show that ThinkFake outperforms state-of-the-art methods on the GenImage benchmark and demonstrates strong zero-shot generalization on the challenging LOKI benchmark.
  arXiv  Detail & Related papers  (2025-09-24T07:34:09Z)
• Diversity Boosts AI-Generated Text Detection [51.56484100374058]
  DivEye is a novel framework that captures how unpredictability fluctuates across a text using surprisal-based features.<n>Our method outperforms existing zero-shot detectors by up to 33.2% and achieves competitive performance with fine-tuned baselines.
  arXiv  Detail & Related papers  (2025-09-23T10:21:22Z)
• Rethinking Evaluation of Infrared Small Target Detection [105.59753496831739]
  This paper introduces a hybrid-level metric incorporating pixel- and target-level performance, proposing a systematic error analysis method, and emphasizing the importance of cross-dataset evaluation.<n>An open-source toolkit has be released to facilitate standardized benchmarking.
  arXiv  Detail & Related papers  (2025-09-21T02:45:07Z)
• Large Language Models for Multilingual Vulnerability Detection: How Far Are We? [13.269680075539135]
  We evaluate the effectiveness of pre-trained language models (PLMs) and large language models (LLMs) for multilingual vulnerability detection.<n>Using over 30,000 real-world vulnerability-fixing patches across seven programming languages, we assess model performance at both the function-level and line-level.<n>Our key findings indicate that GPT-4o, enhanced through instruction tuning and few-shot prompting, significantly outperforms all other evaluated models.
  arXiv  Detail & Related papers  (2025-06-09T07:27:49Z)
• AuthGuard: Generalizable Deepfake Detection via Language Guidance [39.18916434250689]
  Existing deepfake detection techniques struggle to keep-up with the ever-evolving novel, unseen forgeries methods.<n>We propose that incorporating language guidance can improve deepfake detection generalization.<n>We train an expert deepfake vision encoder by combining discriminative classification with image-text contrastive learning.
  arXiv  Detail & Related papers  (2025-06-04T22:50:07Z)
• Reliably Bounding False Positives: A Zero-Shot Machine-Generated Text Detection Framework via Multiscaled Conformal Prediction [18.76704882926895]
  Most existing detection methods focus excessively on detection accuracy, neglecting the societal risks posed by high false positive rates (FPRs)<n>This paper addresses this issue by leveraging Conformal Prediction (CP), which effectively constrains the upper bound of FPRs.<n>To overcome this trade-off, this paper proposes a Zero-Shot Machine-Generated Text Detection Framework via Multiscaled Conformal Prediction (MCP)
  arXiv  Detail & Related papers  (2025-05-08T09:32:38Z)
• Hallucination Detection in LLMs with Topological Divergence on Attention Graphs [64.74977204942199]
  Hallucination, i.e., generating factually incorrect content, remains a critical challenge for large language models.<n>We introduce TOHA, a TOpology-based HAllucination detector in the RAG setting.
  arXiv  Detail & Related papers  (2025-04-14T10:06:27Z)
• Conditioned Prompt-Optimization for Continual Deepfake Detection [11.634681724245933]
  This paper introduces Prompt2Guard, a novel solution for photorealistic-free continual deepfake detection of images. We leverage a prediction ensembling technique with read-only prompts, mitigating the need for multiple forward passes. Our method exploits a text-prompt conditioning tailored to deepfake detection, which we demonstrate is beneficial in our setting.
  arXiv  Detail & Related papers  (2024-07-31T12:22:57Z)
• On the Robustness of Language Guidance for Low-Level Vision Tasks: Findings from Depth Estimation [71.72465617754553]
  We generate "low-level" sentences that convey object-centric, three-dimensional spatial relationships, incorporate them as additional language priors and evaluate their downstream impact on depth estimation. Our key finding is that current language-guided depth estimators perform optimally only with scene-level descriptions. Despite leveraging additional data, these methods are not robust to directed adversarial attacks and decline in performance with an increase in distribution shift.
  arXiv  Detail & Related papers  (2024-04-12T15:35:20Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.