UZH-CL at ArA-DF 2026: Prompt-Tuned Foundation Models and Track-Adaptive Score Fusion for Arabic Speech Deepfake Detection

Authors: Aref Farhadipour, Teodora Vukovic, Petr Motlicek

Published: 2026-10-01 11:56:09+00:00

Comment: this is the submitted system of the UZH-CL team for ArA-DF 2026 challenge

AI Summary

This paper presents the UZH-CL submission to the ArA-DF 2026 Shared Task on Arabic speech deepfake detection, focusing on dialect generalization and acoustic robustness. The approach uses prompt-tuned foundation models, specifically a frozen W2V-BERT-2.0 backbone adapted with Wavelet Prompt Tuning, and track-adaptive score fusion of complementary detectors. This method achieved significant relative reductions in Equal Error Rate (EER) compared to baselines.

Abstract

Detecting synthetic and voice-converted speech remains difficult for low-resource languages with dialectal diversity, where systems must generalize across regional dialects and unseen acoustic channels. We present the UZH-CL submission to the ArA-DF 2026 Shared Task on Arabic speech deepfake detection, covering Track~1 (dialect generalization) and Track~2 (acoustic robustness). We freeze a W2V-BERT-2.0 backbone and adapt it with \\emph{Wavelet Prompt Tuning}, updating under 1% of parameters, and aggregate multi-layer representations with cross-layer attention and a general attentive-statistics pooling head rather than a specialized graph backend. Complementary detectors are obtained by varying adaptation strategy, training data, augmentation, and encoder family. We find that the two shift types require different fusion regimes: a broad multi-window ensemble for dialect generalization, and a compact, channel-matched, center-crop ensemble for acoustic robustness. Official evaluation yields 1.96% EER on Track~1 (6th place) and 1.04% EER on Track~2 (3rd place), corresponding to 87% and 96% relative reductions over the XLS-R+AASIST baselines.