Research map / Vision-language and multimodal / Cross-modal learning
Break Visual-Linguistic Asymmetry: Unleashing VLM's Cross-Modal Potential for General Face Forgery Detection
Guilin Pang, Yiu-ming Cheung, Ruiqi Li, Weifeng Su
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 Meta Forgery Adapter (MFA) (ablation) | CLIP baseline M1 · Celeb-DF-v2 (trained on FF++ c23) | image-level AUC (%): 75.03 -> 87.94 | Improved |
| add Asymmetry-Guided Interactor (AGI) (ablation) | CLIP baseline M1 · Celeb-DF-v2 (trained on FF++ c23) | image-level AUC (%): 75.03 -> 87.92 | Improved |
| add AGI on top of MFA (ablation) | MFA-only model · Celeb-DF-v2 (trained on FF++ c23) | image-level AUC (%): 87.94 -> 92.94 | Improved |
| add MFA on top of AGI (ablation) | AGI-only model · Celeb-DF-v2 (trained on FF++ c23) | image-level AUC (%): 87.92 -> 92.94 | Improved |
| add MFA and AGI (ablation) | CLIP baseline M1 · DFDC-L (trained on FF++ c23) | image-level AUC (%): 72.16 -> 83.60 | Improved |
| add AGI on top of MFA (ablation) | MFA-only model · DFD (trained on FF++ c23) | image-level AUC (%): 92.69 -> 93.85 | Improved |
| add AGI on top of MFA (ablation) | MFA-only model · Celeb-DF-v2 (trained on FF++ c23) | AP (%): 93.76 -> 96.43 | Mixed |
| use asymmetric explicit interaction instead of generic symmetric multi-modal methods (vs MMA) | Asy-Det · Cross-dataset average (DFD, CDF-v2, DFDC-L) | image-level AUC: 0.838 -> 0.901 | Improved |
| replace UDD with Asy-Det | cross-dataset detector · Celeb-DF-v2 (trained on FF++ c23) | image-level AUC: 0.869 -> 0.929 | Improved |
| apply generic symmetric interaction (TCP, CoCoOp) to CLIP | CLIP-B16 · Cross-dataset average (DFD, CDF-v2, DFDC-L) | image-level AUC: 0.775 -> 0.775 | No clear effect |
| Asy-Det vs SSG trained on FaceShifter, tested on DF | Asy-Det · FaceForensics++ cross-manipulation | AUC (%): 94.02 vs 92.22 avg | Mixed |
Closest papers by meaning
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.90)
- A Hitchhiker's Guide to Fine-Grained Face Forgery Detection Using Common Sense Reasoning (NeurIPS 2024 · similarity 0.89)
- VLForgery Face Triad: Detection, Localization and Attribution via Multimodal Large Language Models (NeurIPS 2025 · similarity 0.89)
- Rethinking Vision-Language Model in Face Forensics: Multi-Modal Interpretable Forged Face Detector (CVPR 2025 · similarity 0.89)
- Towards General Visual-Linguistic Face Forgery Detection (CVPR 2025 · similarity 0.89)
- Forensics-Bench: A Comprehensive Forgery Detection Benchmark Suite for Large Vision Language Models (CVPR 2025 · similarity 0.85)
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.
