Get Started
Home
Topics
Search
Library
Research questionHow can audio deepfake detectors identify and localize manipulation when genuine and fake content coexist?A whole-clip label can conceal which time intervals or overlapping sources provide evidence of manipulation. This makes mixed-authenticity audio decisions difficult to interpret and verify.
AI
AI Agents
Audio & Speech
Audio & Speech Processing
Evaluation & Benchmarks
Machine Learning
Reasoning
Research Paper
Small / On-device Models
Sound
Latest papersRecent research connected to this question, newest first.ToolDF: Tool-Integrated Reasoning for Mixed-Authenticity Audio Deepfake DetectionThe source studies mixed-authenticity audio deepfake detection using an audio large language model that can selectively separate sources, consult domain-specific experts, and localize evidence. Its evidence includes a benchmark covering temporal transitions, acoustic overlaps, and hybrid mixtures, along with reported detection and interpretability results.research paper · Sep 7, 2026Half-Truth Audio Detection and Localisation: A Lightweight Cross-Attentive Architecture and a Cross-Corpus Diagnostic StudyThe evidence concerns a lightweight 576K-parameter, 2.24 MB cross-attentive architecture using MFCC, LFCC, and Chroma-STFT features. It is evaluated in-domain on MLADDC T2+T3 and zero-shot on HAD and PartialSpoof: HAD reports 84.9% detection recall and 50.4% correct half-truth classification, while PartialSpoof binary detection is near chance (AUC 0.5544); reported CPU latency is about 14 ms.research paper · Sep 2, 2026
Related questions
How can speech deepfake detectors focus on synthesis artifacts while generalizing to unseen speakers?How can video deepfake detectors adapt to new forgery patterns without forgetting prior spatial and temporal cues?How can watermarks in generated speech remain detectable after neural audio transformations?How can image forensics distinguish real, fully synthetic, and tampered images while localizing altered pixels?