Research map / Label-efficient and robust learning / OOD and anomaly detection
Scaling Behavior in Model Fine-tuning for Audio DeepFake Detection
Xiang Li, Pin-Yu Chen, Wenqi Wei
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 |
|---|---|---|---|
| increase detector model size (Whisper Tiny to Large) (ablation) | Whisper LoRA deepfake detector · in-distribution | EER | Improved |
| increase fine-tuning data fraction (2% to 100%) (ablation) | Whisper LoRA deepfake detector · in-distribution | EER | Improved |
| scale model and data under out-of-distribution evaluation (ablation) | Whisper LoRA deepfake detector · OOD benchmarks (In-the-Wild, ADD, ASVspoof2021 etc.) | EER | Mixed |
| increase model size under Gaussian noise corruption (ablation) | Whisper LoRA deepfake detector · Gaussian noise | EER | Improved |
| increase model size under pitch shifting (ablation) | Whisper LoRA deepfake detector · pitch shift corruption | EER | Mixed |
| increase model size under Encodec distortion (ablation) | Whisper LoRA deepfake detector · Encodec neural codec corruption | EER | No clear effect |
| increase model size with limited training compute (ablation) | Whisper LoRA deepfake detector · compute-matched fine-tuning | EER | Hurt |
| increase model size with large training compute (ablation) | Whisper LoRA deepfake detector · compute-matched fine-tuning | EER | Improved |
| increase model size for English-only trained detector on unseen languages (ablation) | Whisper LoRA deepfake detector · SpeechFake (46 non-English languages) | EER | Improved |
| increase model size on unseen TTS systems (ablation) | Whisper LoRA deepfake detector · cross-TTS generalization | EER | Improved |
Closest papers by meaning
- Adaptive Compression and Targeted Perturbation: A Unified Framework for Generalized Audio Deepfake Detection (NeurIPS 2026 · similarity 0.90)
- I Can Hear You: Selective Robust Training for Deepfake Audio Detection (ICLR 2025 · similarity 0.89)
- Region-Based Optimization in Continual Learning for Audio Deepfake Detection (AAAI 2025 · similarity 0.88)
- Circumventing Shortcuts in Audio-visual Deepfake Detection Datasets with Unsupervised Learning (CVPR 2025 · similarity 0.87)
- What to Remember: Self-Adaptive Continual Learning for Audio Deepfake Detection (AAAI 2024 · similarity 0.87)
- Phoneme-Level Feature Discrepancies: A Key to Detecting Sophisticated Speech Deepfakes (AAAI 2025 · similarity 0.86)
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.
