Research map / Label-efficient and robust learning / Domain adaptation and generalization
Suppressing Gradient Conflict for Generalizable Deepfake Detection
Ming-Hui Liu, Harry Hui Cheng, Xin Luo, X.G Xu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add UVS module (ablation) | EfficientNet backbone · Celeb-DF | AUC (%): 64.59 -> 94.45 | Improved |
| add CGR module (ablation) | EfficientNet backbone · Celeb-DF | AUC (%): 64.59 -> 94.92 | Improved |
| add CGR on top of UVS (ablation) | CS-DFD · Celeb-DF | AUC (%): 94.45 -> 95.32 | Improved |
| add CGR on top of UVS (ablation) | CS-DFD · DFDC | AUC (%): 73.65 -> 74.97 | Improved |
| add CGR on top of UVS (ablation) | CS-DFD · FF++ (test) | AUC (%): 97.98 -> 98.88 | Improved |
| add online synthesized forgeries to original-only training | Xception · FF++ (test) | AUC (%): 96.37 -> 94.86 | Hurt |
| add online synthesized forgeries to original-only training | Xception · Celeb-DF | AUC (%): 56.75 -> 87.77 | Improved |
| add original forgeries to synthesized-only training (1+1<2) | Xception · Celeb-DF | AUC (%): 90.11 -> 87.77 | Hurt |
| train on synthesized forgeries instead of original | ViT-L · FF++ (test) | AUC (%): 97.65 -> 70.32 | Hurt |
| apply CS-DFD instead of naive O+S training | Xception · FF++ (test) | AUC (%): 94.86 -> 97.48 | Improved |
| apply CS-DFD instead of naive O+S training | ViT-L · Celeb-DF | AUC (%): 85.15 -> 90.26 | Improved |
| apply CS-DFD instead of naive O+S training | ViT-B · Celeb-DF | AUC (%): 91.28 -> 93.66 | Improved |
Closest papers by meaning
- FreqDebias: Towards Generalizable Deepfake Detection via Consistency-Driven Frequency Debiasing (CVPR 2025 · similarity 0.89)
- Exploring Unbiased Deepfake Detection via Token-Level Shuffling and Mixing (AAAI 2025 · similarity 0.89)
- D^3: Scaling Up Deepfake Detection by Learning from Discrepancy (CVPR 2025 · similarity 0.88)
- Transcending Forgery Specificity with Latent Space Augmentation for Generalizable Deepfake Detection (CVPR 2024 · similarity 0.88)
- Veritas: Generalizable Deepfake Detection via Pattern-Aware Reasoning (ICLR 2026 · similarity 0.88)
- Exposing the Deception: Uncovering More Forgery Clues for Deepfake Detection (AAAI 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
