Wanying Ge

A training-free self-embedding strategy can significantly enhance the detection of partial deepfake speech, even when only small segments are manipulated.

Artifact-specific experts can provide interpretable evidence in speech deepfake detection, transforming how we assess synthetic audio authenticity.

RL fine-tuning with GRPO can significantly boost the generalization of speech deepfake detectors to unseen attacks, outperforming standard supervised methods.

Repurpose Whisper as a deepfake detector and you can transcribe and spot synthetic words in one pass, rivaling dedicated models.