VIGIL: Part-Grounded Structured Reasoning for Generalizable Deepfake Detection

TL;DR

VIGIL introduces a part-grounded, structured reasoning framework for deepfake detection that improves interpretability and generalizability by mimicking forensic expert practices and employing a multi-stage training paradigm.

Contribution

The paper presents VIGIL, a novel forensic framework with a plan-then-examine pipeline, part-aware evidence integration, and a hierarchical benchmark for robust generalization in deepfake detection.

Findings

VIGIL outperforms existing methods on OmniFake benchmark.

The framework achieves high accuracy across diverse datasets.

Part-aware training enhances generalizability and interpretability.

Abstract

Multimodal large language models (MLLMs) offer a promising path toward interpretable deepfake detection by generating textual explanations. However, the reasoning process of current MLLM-based methods combines evidence generation and manipulation localization into a unified step. This combination blurs the boundary between faithful observations and hallucinated explanations, leading to unreliable conclusions. Building on this, we present VIGIL, a part-centric structured forensic framework inspired by expert forensic practice through a plan-then-examine pipeline: the model first plans which facial parts warrant inspection based on global visual cues, then examines each part with independently sourced forensic evidence. A stage-gated injection mechanism delivers part-level forensic evidence only during examination, ensuring that part selection remains driven by the model's own perception rather than biased by external signals. We further propose a progressive three-stage training paradigm whose reinforcement learning stage employs part-aware rewards to enforce anatomical validity and evidence--conclusion coherence. To enable rigorous generalizability evaluation, we construct OmniFake, a hierarchical 5-Level benchmark where the model, trained on only three foundational generators, is progressively tested up to in-the-wild social-media data. Extensive experiments on OmniFake and cross-dataset evaluations demonstrate that VIGIL consistently outperforms both expert detectors and concurrent MLLM-based methods across all generalizability levels.