# VIGIL: Part-Grounded Structured Reasoning for Generalizable Deepfake Detection Source: [VIGIL: Part-Grounded Structured Reasoning for Generalizable Deepfake Detection]() Collection index: [llms.txt]() Data updated: 2026-07-01T05:00:12.177Z Authors: Xinghan Li, Junhao Xu, Jingjing Chen Published: 2026-03-23T03:35:36.000Z ## 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. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Deepfake Detection on OmniFake (In-domain)](): Deepfake Detection; OmniFake (In-domain); ADM Accuracy: 98.6. - [Deepfake Detection on OmniFake Cross-Model](): Deepfake Detection; OmniFake Cross-Model; Midjourney Accuracy: 99.4. - [Deepfake Detection on OmniFake Cross-Task](): Deepfake Detection; OmniFake Cross-Task; Accuracy (BrushNet): 92. - [Deepfake Detection on OmniFake (In-the-Wild)](): Deepfake Detection; OmniFake (In-the-Wild); DFDC Accuracy: 88. - [Deepfake Detection on OmniFake Full Suite](): Deepfake Detection; OmniFake Full Suite; Accuracy: 93.1. - [Deepfake Detection on HydraFake](): Deepfake Detection; HydraFake; ID Accuracy: 98.2. - [Deepfake Detection on OmniFake Cross-Architecture](): Deepfake Detection; OmniFake Cross-Architecture; Accuracy (FF++): 92.8. ## Original sources and implementation [Original paper]() [PDF]()