Similarity Choice and Negative Scaling in Supervised Contrastive Learning for Deepfake Audio Detection
This work provides a controlled comparison of SupCon variants for the specific problem of audio deepfake detection, offering practical insights but is incremental in nature.
The paper studies supervised contrastive learning (SupCon) for audio deepfake detection, finding that cosine similarity with a delayed negative queue achieves the best in-the-wild EER (8.29%) and pooled EER (4.44), while angular similarity performs well without large negative sets (ITW 8.70).
Supervised contrastive learning (SupCon) is widely used to shape representations, but has seen limited targeted study for audio deepfake detection. Existing work typically combines contrastive terms with broader pipelines; however, the focus on SupCon itself is missing. In this work, we run a controlled study on wav2vec2 XLS-R (300M) that varies (i) similarity in SupCon (cosine vs angular similarity derived from the hyperspherical angle) and (ii) negative scaling using a warm-started global cross-batch queue. Stage 1 fine-tunes the encoder and projection head with SupCon; Stage 2 freezes them and trains a linear classifier with BCE. Trained on ASVspoof 2019 LA and evaluated on ASV19 eval plus ITW and ASVspoof 2021 DF/LA, Cosine SupCon with a delayed queue achieves the best ITW EER (8.29%) and pooled EER (4.44), while angular similarity performs strongly without queued negatives (ITW 8.70), indicating reduced reliance on large negative sets.