Research map / Detection and segmentation / Open-set and few-shot detection
Towards Understanding the Generalization of Deepfake Detectors from a Game-Theoretical View
Kelu Yao, Jin Wang, Boyu Diao, Chao Li
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 |
|---|---|---|---|
| subtract the low-order-interaction part of the output score at inference (Eq. 8) | ResNet-18 without augmentation · Celeb-DF v1 | F-AUC (%): 56.76 -> 57.56 | Improved |
| subtract the low-order-interaction part of the output score at inference (Eq. 8) | ResNet-18 with augmentation · Celeb-DF v1 | F-AUC (%): 70.55 -> 70.52 | No clear effect |
| subtract the low-order-interaction part of the output score at inference (Eq. 8) | SOTA detectors (SBI, FST-Matching, CADDM) · Celeb-DF v1 | F-AUC (%): SBI 85.21 -> 85.23; CADDM 80.24 -> 80.27 | No clear effect |
| subtract the low-order-interaction part of the output score at inference (Eq. 8) | ResNet-18 without augmentation · FF++ in-dataset | F-AUC (%): 99.61 -> 99.60 | No clear effect |
| train with data augmentation (color jitter, crop, blur/noise, JPEG) (ablation) | ResNet-34 · Celeb-DF v1 | F-AUC (%): 47.81 -> 81.01 | Improved |
| train with data augmentation reduces low-order interaction strength (ablation) | ResNet-34 · FF++ test, Deepfakes | low-order interaction strength rho: 0.102 -> 0.068 | Improved |
| train with data augmentation on Efficient-b3 low-order interaction strength (ablation) | Efficient-b3 · FF++ test, Deepfakes | low-order interaction strength rho: 0.050 -> 0.052 | No clear effect |
Closest papers by meaning
- D^3: Scaling Up Deepfake Detection by Learning from Discrepancy (CVPR 2025 · similarity 0.86)
- Exploring Unbiased Deepfake Detection via Token-Level Shuffling and Mixing (AAAI 2025 · similarity 0.86)
- From Specificity to Generality: Revisiting Generalizable Artifacts in Detecting Face Deepfakes (NeurIPS 2025 · similarity 0.85)
- Standing on the Shoulders of Giants: Reprogramming Visual-Language Model for General Deepfake Detection (AAAI 2025 · similarity 0.85)
- PRPO: Paragraph-level Policy Optimization for Vision-Language Deepfake Detection (ICML 2026 · similarity 0.84)
- Fair Deepfake Detectors Can Generalize (NeurIPS 2025 · similarity 0.84)
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.
