论文解读
流式多说话人识别要在单实例省内存与多实例保重叠之间选边
该文用同一组流式识别与流式说话人日志模型派生出级联、掩码输入、词级串行输出和日志条件四种架构，对比多说话人准确率、单说话人退化、内存占用与训练代价，并以排列不变动态时间规整解决短片段训练标签对齐问题，其中日志条件精度最好但需要微调与多实例代价。
模型融合
子标签
下方只统计已核验方向的论文记录，按已核身份去重，身份待核时按页面保留。“直接”只含明确标注该方向的论文；“含子树”同时包含下级方向。点击名称进入论文库继续筛选；数量与上方按标签名称收录的页面数不同。
- 模型集成 直接 84 篇 · 含子树 84 篇
论文解读
不训练也能听懂说话：冻结视觉语言模型的外挂语音路由何时够用
论文用冻结视觉语言模型加 Whisper 转写路由构造免训练语音中心全模态理解，在 56 个基准和 21 种语言的配对比较中语音任务可比原生全模态训练而保留视觉推理能力，代价是增加串行语音识别延迟且无法处理音乐与环境声等非语音线索。
论文解读
当嵌入不再确定：把不确定性从打分一路带到校准的后端闭环
说话人验证 | 7.8/10
论文解读
类型未知时如何判真假：用两套耳朵听同一段音频
音频伪造检测 | 6.3/10
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
SonarLLM: A Native Sonar--Optical Multimodal Large Language Model for Underwater Perception
音频理解 | 8.0/10
论文解读
Adaptive Hierarchical Representation Alliance for Multimodal Learning
语音情感识别 | 8.6/10
论文解读
Mitigating Speaker Leakage in Cascaded Multi-talker ASR with Diarization-based Transcript Correction
语音识别 | 7.1/10
论文解读
Multi-Modal Semantic Expansion with Constrained LLM Reranking for Conversational Music Recommendation
音乐推荐 | 8.3/10
论文解读
TurboBias 2.0: Streaming Context-Biasing for Production-Efficient ASR Systems
语音识别 | 9.0/10
论文解读
Multimodal Rapport Estimation in Real-World HRI
多模态模型 | 7.0/10
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
Smartphone Audio Based Distress Detection
音频事件检测 | 6.3/10
论文解读
Music-Source-Separation-Training (MSST): A Unified Framework for Training and Evaluating Music Demixing Models
音乐源分离 | 5.7/10
论文解读
AlphaWiSE: Adaptive Weight Interpolation for Continual Multimodal Representation Learning
音频检索 | 6.4/10
论文解读
Unlocking Speech–Text Compositional Powers: Instruction-Following Speech Language Models without Instruction Tuning
语音交互 | 6.7/10
论文解读
Unlocking Speech-Text Compositional Powers: Instruction-Following Speech Language Models without Instruction Tuning
语音交互 | 8.5/10
论文解读
MuVAP: Multimodal Multiparty Voice Activity Projection for Turn-taking Prediction in the Wild
语音对话系统 | 7.8/10
论文解读
Stabilizing Short Duration Speaker Verification through Neural Re-scoring with Hybrid Enrollment
说话人验证 | 7.9/10
论文解读
Conan-embedding-v3: Fusing Modality-Specific Models for Omni-Modal Embedding
音频检索 | 7.6/10