# Deepfake Detection on OmniFake Cross-Task Source: [Deepfake Detection on OmniFake Cross-Task]() Collection index: [llms.txt]() Data updated: 2026-03-24T15:42:27.576Z Evaluation of Deepfake Detection models on the OmniFake Cross-Task dataset, reporting accuracy metrics specific to various source deepfake generation methods (BrushNet, CodeF., GFPGAN, RestoreF., Chameleon). 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. | Scope | Recorded value | | --- | --- | | Task | Deepfake Detection | | Dataset (including any stated split) | OmniFake Cross-Task | | Metric | Accuracy (BrushNet) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 17 of 17 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Accuracy (BrushNet)](): Higher is better; unit: PERCENT. - [Accuracy (CodeF.)](): Higher is better; unit: PERCENT. - [Accuracy (GFPGAN)](): Higher is better; unit: PERCENT. - [Accuracy (RestoreF.)](): Higher is better; unit: PERCENT. - [Accuracy (Chameleon)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy (BrushNet) | Result date | Report paper | | --- | --- | --- | --- | | VIGIL | 92 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | Co-SPY | 88.4 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | DDA | 86.7 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | VIGIL Cold-Start | 85.4 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | Veritas | 84 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | D^3 | 73.8 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | AIDE | 73.5 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | FakeVLM | 72.8 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | UnivFD | 72 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | Gemini-3-Pro | 68.2 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | NPR | 67.6 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | Kimi-K2.5 | 65.2 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | C2P-CLIP | 63.4 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | GPT-5.2 | 63.4 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | GLM-4.6V | 54.1 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | SAFE | 50.4 | Not recorded | 69c21a9f0636e5bf4afc49c3 | | Qwen3-VL-8B | 48.9 | Not recorded | 69c21a9f0636e5bf4afc49c3 | ### VIGIL Accuracy (BrushNet): 92. Result ID: 69c2b0d147f34ce924abd787. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-VL-8B"}. This record does not assert independent reproduction. ### Co-SPY Accuracy (BrushNet): 88.4. Result ID: 69c2b0d147f34ce924abd77d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### DDA Accuracy (BrushNet): 86.7. Result ID: 69c2b0d147f34ce924abd77e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### VIGIL Cold-Start Accuracy (BrushNet): 85.4. Result ID: 69c2b0d147f34ce924abd786. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-VL-8B","mode":"Cold-Start"}. This record does not assert independent reproduction. ### Veritas Accuracy (BrushNet): 84. Result ID: 69c2b0d147f34ce924abd785. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### D^3 Accuracy (BrushNet): 73.8. Result ID: 69c2b0d147f34ce924abd77c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### AIDE Accuracy (BrushNet): 73.5. Result ID: 69c2b0d147f34ce924abd77a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### FakeVLM Accuracy (BrushNet): 72.8. Result ID: 69c2b0d147f34ce924abd784. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### UnivFD Accuracy (BrushNet): 72. Result ID: 69c2b0d147f34ce924abd777. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Gemini-3-Pro Accuracy (BrushNet): 68.2. Result ID: 69c2b0d147f34ce924abd783. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### NPR Accuracy (BrushNet): 67.6. Result ID: 69c2b0d147f34ce924abd778. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Kimi-K2.5 Accuracy (BrushNet): 65.2. Result ID: 69c2b0d147f34ce924abd781. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### C2P-CLIP Accuracy (BrushNet): 63.4. Result ID: 69c2b0d147f34ce924abd77b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### GPT-5.2 Accuracy (BrushNet): 63.4. Result ID: 69c2b0d147f34ce924abd782. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### GLM-4.6V Accuracy (BrushNet): 54.1. Result ID: 69c2b0d147f34ce924abd780. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### SAFE Accuracy (BrushNet): 50.4. Result ID: 69c2b0d147f34ce924abd779. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Qwen3-VL-8B Accuracy (BrushNet): 48.9. Result ID: 69c2b0d147f34ce924abd77f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"zero-shot":"true"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.