# Enhancing Self-Supervised Talking Head Forgery Detection via a Training-Free Dual-System Framework

arXiv: https://arxiv.org/abs/2605.03390 · HTML: https://mlsanity.com/paper/2605.03390/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 24 of its 50 references that are in the map.
- First posted on arXiv 5 May 2026.

## Builds on (most cited first, 24 of 24)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [CLIP-Adapter: Better Vision-Language Models with Feature Adapters](https://mlsanity.com/paper/2110.04544/) (2021)
- [Learning Audio-Visual Speech Representation by Masked Multimodal Cluster Prediction](https://mlsanity.com/paper/2201.02184/) (2022)
- [Efficient Test-Time Adaptation of Vision-Language Models](https://mlsanity.com/paper/2403.18293/) (2024)
- [Leveraging Real Talking Faces via Self-Supervision for Robust Forgery Detection](https://mlsanity.com/paper/2201.07131/) (2022)
- [AEROBLADE: Training-Free Detection of Latent Diffusion Images Using Autoencoder Reconstruction Error](https://mlsanity.com/paper/2401.17879/) (2024)
- [Emotions Don't Lie: An Audio-Visual Deepfake Detection Method Using Affective Cues](https://mlsanity.com/paper/2003.06711/) (2020)
- [SIDA: Social Media Image Deepfake Detection, Localization and Explanation with Large Multimodal Model](https://mlsanity.com/paper/2412.04292/) (2024)
- [Not made for each other- Audio-Visual Dissonance-based Deepfake Detection and Localization](https://mlsanity.com/paper/2005.14405/) (2020)
- [SyncTalk: The Devil is in the Synchronization for Talking Head Synthesis](https://mlsanity.com/paper/2311.17590/) (2023)
- [Self-Supervised Video Forensics by Audio-Visual Anomaly Detection](https://mlsanity.com/paper/2301.01767/) (2023)
- [Combining EfficientNet and Vision Transformers for Video Deepfake Detection](https://mlsanity.com/paper/2107.02612/) (2021)
- [Deepfake Video Detection Using Convolutional Vision Transformer](https://mlsanity.com/paper/2102.11126/) (2021)
- [Audio-Visual Synchronisation in the wild](https://mlsanity.com/paper/2112.04432/) (2021)
- [LampMark: Proactive Deepfake Detection via Training-Free Landmark Perceptual Watermarks](https://mlsanity.com/paper/2411.17209/) (2024)
- [ProAPO: Progressively Automatic Prompt Optimization for Visual Classification](https://mlsanity.com/paper/2502.19844/) (2025)
- [Cafe-Talk: Generating 3D Talking Face Animation with Multimodal Coarse- and Fine-grained Control](https://mlsanity.com/paper/2503.14517/) (2025)
- [Zero-Shot Fake Video Detection by Audio-Visual Consistency](https://mlsanity.com/paper/2406.07854/) (2024)
- [TalkingHeadBench: A Multi-Modal Benchmark & Analysis of Talking-Head DeepFake Detection](https://mlsanity.com/paper/2505.24866/) (2025)
- [Training-free Detection of AI-generated images via Cropping Robustness](https://mlsanity.com/paper/2511.14030/) (2025)
- [LLMs as Bridges: Reformulating Grounded Multimodal Named Entity Recognition](https://mlsanity.com/paper/2402.09989/) (2024)
- [KLASSify to Verify: Audio-Visual Deepfake Detection Using SSL-based Audio and Handcrafted Visual Features](https://mlsanity.com/paper/2508.07337/) (2025)
- [Global Prompt Refinement with Non-Interfering Attention Masking for One-Shot Federated Learning](https://mlsanity.com/paper/2509.22700/) (2025)

## Abstract (arXiv)

Supervised talking head forgery detection faces severe generalization challenges due to the continuous evolution of generators. By reducing reliance on generator-specific forgery patterns, self-supervised detectors offer stronger cross-generator robustness. However, existing research has mainly focused on building stronger detectors, while the discriminative capacity of trained detectors remains insufficiently exploited. In particular, for score-based self-supervised detectors, the limited discriminative ability on hard cases is often reflected in unreliable anomaly ordering, leaving room for further refinement. Motivated by this observation, we draw inspiration from the dual-system theory of human cognition and propose a Training-Free Dual-System (TFDS) framework to further exploit the latent discriminative capacity of existing score-based self-supervised detectors. TFDS treats anomaly-like scores as the basis of System-1, using lightweight threshold-based routing to partition samples into confident and uncertain subsets. System-2 then revisits only the uncertain subset, performing fine-grained evidence-guided reasoning to refine the relative ordering of ambiguous samples within the original score distribution. Extensive experiments demonstrate consistent improvements across datasets and perturbation settings, with the gains arising mainly from corrected ordering within the uncertain subset. These findings show that existing self-supervised talking head forgery detectors still contain underexploited discriminative cues that can be effectively unlocked through training-free dual-system reasoning.