Research map / Label-efficient and robust learning / Recognition and long-tail
BPL: Generalizable Deepfake Detection via Bias-only Pair-aware Learning
Yuxiang Xu, Rundong He, Zhiyuan Yan, Yicong Dong, Zhongyi Han, Xiaoyan Wang, Yilong Yin
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 |
|---|---|---|---|
| use bias-only tuning instead of full fine-tuning of the backbone (ablation) | CLIP ViT-L/14 deepfake detector · DF40 cross-manipulation | average AUC (%): 43.16 -> 76.22 | Improved |
| add paired real-fake data sampling (ablation) | bias-only CLIP detector · DF40 cross-manipulation | average AUC (%): 76.22 -> 81.81 | Improved |
| add pair-wise discrepancy (contrastive) loss (ablation) | BPL · DF40 cross-manipulation | average AUC (%): 81.81 -> 88.86 | Improved |
| use paired training data instead of all unpaired fake frames (ablation) | BPL · DF40 (SimSwap, e4s, InSwap, DFLab) | average AUC (%): 80.20 -> 81.12 | Improved |
| use paired training data instead of all unpaired fake frames (ablation) | Effort · DF40 (SimSwap, e4s, InSwap, DFLab) | average AUC (%): 79.63 -> 78.64 | Hurt |
| reduce training data from 100% to 10% (ablation) | BPL · AP on DFDC, DFDCP, SimSwap, FSGAN | average AP (%): 90.34 -> 88.99 | Hurt |
| reduce training data from 100% to 10% (ablation) | Effort · AP on DFDC, DFDCP, SimSwap, FSGAN | average AP (%): 85.39 -> 81.13 | Hurt |
| vary loss weight lambda in 0.05, 0.4, 0.7 (ablation) | BPL · hyperparameter sensitivity | AP (%): peak 95.43 at lambda=0.4 | No clear effect |
| use overly large margin m (ablation) | BPL · hyperparameter sensitivity | AP (%) | Hurt |
Closest papers by meaning
- D^3: Scaling Up Deepfake Detection by Learning from Discrepancy (CVPR 2025 · similarity 0.89)
- DiffusionFake: Enhancing Generalization in Deepfake Detection via Guided Stable Diffusion (NeurIPS 2024 · similarity 0.88)
- FreqBlender: Enhancing DeepFake Detection by Blending Frequency Knowledge (NeurIPS 2024 · similarity 0.88)
- Towards More General Video-based Deepfake Detection through Facial Component Guided Adaptation for Foundation Model (CVPR 2025 · similarity 0.88)
- Standing on the Shoulders of Giants: Reprogramming Visual-Language Model for General Deepfake Detection (AAAI 2025 · similarity 0.88)
- From Specificity to Generality: Revisiting Generalizable Artifacts in Detecting Face Deepfakes (NeurIPS 2025 · 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.
