论文解读
加了编解码器训练就掉音质：CRAW 用注意力、掩蔽与纠错把两者拉回平衡
CRAW 针对神经编解码器与去噪重合成导致后处理水印失效的问题，用加宽失真训练加注意力池化加推理时感知掩蔽加重复纠错码，在保持 PESQ 约 4.0 的同时把编解码器与去噪下的检测 F1 拉回 0.8 以上，代价是掩蔽带来约 150 ms 的额外推理延迟。
语音伪造检测
论文解读
真假混在一起时：让大模型先分轨再判真伪的 ToolDF
针对一段音频里同时含真实与伪造、跨时间与跨声源的混合真实性检测，ToolDF 用音频大模型做编排器按结构调用分离与四个领域专家并执行早失败汇总，在复合类型上取得最高的 C-Avg. 81.89，代价是依赖外部工具的可靠性。
论文解读
听不见的指纹更可信？把语音伪造溯源拆成可感知与不可感知两条线索来验
语音伪造检测 | 7.4/10
论文解读
别再只调大模型：用 88 行文字把声音的破绽说给 LLM 听
语音伪造检测 | 6.4/10
论文解读
别让检测器猜伪迹：先把原声的压缩记忆藏回去
语音伪造检测 | 6.5/10
论文解读
想找回被替换的原话，先接受水印不能再像哈希那样精确
音频水印 | 6.4/10
论文解读
Investigating voiced and unvoiced regions of speech for audio deepfake detection
语音伪造检测 | 6.4/10
论文解读
Explainability by Design: Structured Kolmogorov-Arnold Networks over Probabilistic Attributes for Speech Deepfake Source Tracing
语音伪造检测 | 7.5/10
论文解读
Tracking the Trend in How Speech Synthesizers Deceive People
语音伪造检测 | 6.0/10
论文解读
The Last Mile of Deepfake Speech Detection: An Industry-Academia Experience Report
语音伪造检测 | 6.6/10
论文解读
A survey of AI-generated voices and their detection
语音伪造检测 | 6.1/10
论文解读
AT-ADD: All-Type Audio Deepfake Detection Challenge Summary
音频伪造检测 | 6.3/10
论文解读
Trajectory Dynamics in Self-Supervised Learning Latent Space for Audio Deepfake Detection
语音伪造检测 | 7.6/10
论文解读
Evaluating Pre-trained Speech Encoders for Spontaneous Speech Detection and Out of Domain Synthetic Speech Generalisation in Indic Languages
语音伪造检测 | 6.0/10
论文解读
The Voiceprint Fallacy: Why Voices Are Not Unique Biometric Imprints
说话人验证 | 6.0/10
论文解读
AffectDF: The Most Comprehensive Benchmark for Speech Deepfake Detection against Emotionally Expressive Attacks
语音伪造检测 | 8.4/10
论文解读
REIMU: Efficient Heterogeneous Hierarchical Reasoning for SSL-Based Speech Deepfake Detection
语音伪造检测 | 6.4/10
论文解读
Cloned Voices, Real Consequences: Evaluating Bias in Political Deepfake Detection for Electoral Integrity in Brazil
语音伪造检测 | 7.7/10
论文解读
Teffic-Audio: Tell Fact from Fiction
语音伪造检测 | 6.8/10
论文解读
Audio-Anchored Fusion of Multi-Ratio DiT Reconstruction Residuals for Cross-Domain Audio Deepfake Detection
语音伪造检测 | 6.6/10