Research map / Reinforcement learning / Visual and representation RL
Tutor-Student Reinforcement Learning: A Dynamic Curriculum for Robust Deepfake Detection
Zhanhe Lei, Zhongyuan Wang, Jikang Cheng, Baojin Huang, Yuhong Yang, Zhen Han, Chao Liang, Dengpan Ye
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 TSRL tutor | Effort · Cross-dataset (FF++ c23 train; CDF-v2, DFD, DFDC, DFDCP) | avg AUC: 0.886 -> 0.903 | Improved |
| add TSRL tutor | Effort · Cross-method (DF40, FF++ c23 train) | avg AUC: 0.920 -> 0.942 | Improved |
| add TSRL tutor | CLIP · Cross-dataset (FF++ c23 train; CDF-v2, DFD, DFDC, DFDCP) | avg AUC: 0.732 -> 0.768 | Improved |
| add TSRL tutor | CORE · Cross-method (DF40, FF++ c23 train) | avg AUC: 0.814 -> 0.855 | Improved |
| add TSRL tutor | IID · Cross-dataset (FF++ c23 train; CDF-v2, DFD, DFDC, DFDCP) | avg AUC: 0.767 -> 0.784 | Improved |
| add TSRL tutor | ProDet · Cross-method (DF40, FF++ c23 train) | avg AUC: 0.839 -> 0.850 | Improved |
| add TSRL tutor | UCF · Cross-dataset (FF++ c23 train; CDF-v2, DFD, DFDC, DFDCP) | avg AUC: 0.771 -> 0.782 | Improved |
| add TSRL tutor | CORE · Cross-method (DF40, FF++ c23 train) | e4s AUC: 0.658 -> 0.537 | Hurt |
| add static curriculum heuristic (CL) (ablation) | CORE · Cross-method DF40 ablation (CORE) | avg AUC: 0.814 -> 0.817 | No clear effect |
| replace static CL with PPO-trained tutor (ablation) | CORE · Cross-method DF40 ablation (CORE) | avg AUC: 0.817 -> 0.855 | Improved |
| add TSRL tutor | CLIP · Cross-method (DF40, FF++ c23 train) | avg AUC: 0.653 -> 0.674 | Improved |
Closest papers by meaning
- Unsupervised Training Sequence Design: Efficient and Generalizable Agent Training (AAAI 2024 · similarity 0.85)
- PRPO: Paragraph-level Policy Optimization for Vision-Language Deepfake Detection (ICML 2026 · similarity 0.84)
- Deep Demonstration Tracing: Learning Generalizable Imitator Policy for Runtime Imitation from a Single Demonstration (ICML 2024 · similarity 0.84)
- IFDECORATOR: Wrapping Instruction Following Reinforcement Learning with Verifiable Rewards (NeurIPS 2026 · similarity 0.84)
- D^3: Scaling Up Deepfake Detection by Learning from Discrepancy (CVPR 2025 · similarity 0.84)
- LLM-as-a-Tutor: Policy-Aware Prompt Adaptation for Non-Verifiable RL (NeurIPS 2026 · 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.
