论文解读
检索增强让高层意图变低层声学：用锚点保检索、用描述做转向的标题投毒
针对检索增强的文本到音乐生成中用检索到的标题改写用户高层提示的依赖，论文提出保留高层锚点并注入低层声学描述的双层标题投毒，使投毒后生成音频与攻击目标类别的 CLAP 相似度从 0.21-0.28 提升到 0.41-0.48，同时与原始用户问题的相似度维持在约 0.30 附近。
检索增强
论文解读
Bypassing Direct Reconstruction: Speech Detection from MEG via Large-Scale Audio Retrieval
语音活动检测 | 7.0/10
论文解读
A Decomposed Retrieval-Edit-Rerank Framework for Chord Generation
A Decomposed Retrieval-Edit-Rerank Framework for Chord Generation
论文解读
Unified Multi-Modal Interactive and Reactive 3D Motion Generation via Rectified Flow
动作生成 | 7.5/10
论文解读
Feedback-Driven Retrieval-Augmented Audio Generation with Large Audio Language Models
音频生成 | 6.5/10
论文解读
Language-Infused Retrieval-Augmented CTC with Adaptive Soft-Hard Gating for Robust Code-Switching ASR
语音识别 | 8.0/10
论文解读
Mispronunciation Detection and Diagnosis Without Model Training: A Retrieval-Based Approach
语音评估 | 8.0/10
论文解读
ICLAD: In-Context Learning with Comparison-Guidance for Audio Deepfake Detection
本文针对音频深度伪造检测模型在真实场景（in-the-wild）中泛化能力差的核心问题，提出了一种名为ICLAD的全新范式。该框架利用音频语言模型（ALM）的上下文学习能力，实现了无需训练的快速适应。其核心是创新的**成对比较推理**策略：在离线阶段，引导ALM为每个样本同时生成“真实”和“伪造”的