论文解读
MSF-SER: Enriching Acoustic Modeling with Multi-Granularity Semantics for Speech Emotion Recognition
语音情感识别 | 7.5/10
音频大模型
论文解读
Multimodal LLMs as Expert Speech Annotators: Acoustic Macro-Descriptors for Parkinson's Detection
语音生物标志物 | 6.5/10
论文解读
Scaling Ambiguity: Augmenting Human Annotation in Speech Emotion Recognition with Audio-Language Models
语音情感识别 | 6.5/10
论文解读
Segmentwise Pruning in Audio-Language Models
音频问答 | 7.0/10
论文解读
SONAR: Self-Distilled Continual Pre-Training for Domain Adaptive Audio Representation
音频事件检测 | 7.0/10
论文解读
Sparse Autoencoders Make Audio Foundation Models More Explainable
模型评估 | 6.5/10
论文解读
Teaching Audio Models to Reason: A Unified Framework for Source- and Layer-Wise Distillation
音频问答 | 7.0/10
论文解读
Temporal Distillation for Music Representation Learning
音乐信息检索 | 7.5/10
论文解读
Test-Time Scaling for Auditory Cognition in Audio Language Models
音频问答 | 7.0/10
论文解读
The Muse Benchmark: Probing Music Perception and Auditory Relational Reasoning in Audio LLMs
音乐理解 | 8.5/10
论文解读
Walking Through Uncertainty: An Empirical Study of Uncertainty Estimation for Audio-Aware Large Language Models
音频问答 | 7.5/10
论文解读
When Noise Lowers the Loss: Rethinking Likelihood-Based Evaluation in Music Large Language Models
音乐生成 | 7.0/10
论文解读
When Silence Matters: The Impact of Irrelevant Audio on Text Reasoning in Large Audio-Language Models
模型评估 | 7.0/10
论文解读
When Voice Matters: A Controlled Study of Audio LLM Behavior in Clinical Decision-Making
模型评估 | 7.0/10
论文解读
All That Glitters Is Not Audio: Rethinking Text Priors and Audio Reliance in Audio-Language Evaluation
音频问答 | 6.5/10
论文解读
HeadRouter: Dynamic Head-Weight Routing for Task-Adaptive Audio Token Pruning in Large Audio Language Models
音频大模型 | 8.0/10
论文解读
Listening with Time: Precise Temporal Awareness for Long-Form Audio Understanding
音频场景理解 | 8.0/10
论文解读
Benign Fine-Tuning Breaks Safety Alignment in Audio LLMs
这篇论文首次系统研究了良性（无害）音频数据微调对音频大模型安全对齐的破坏作用。**要解决的问题**是：用户出于提升模型性能目的进行的常规微调，是否会无意中破坏模型的安全防护？**方法**上，作者提出了一个基于嵌入空间邻近度的过滤框架，从语义、声学及混合维度，选择性地用与有害内容在表示空间上相近的良性
论文解读
HalluAudio: A Comprehensive Benchmark for Hallucination Detection in Large Audio-Language Models
这篇论文旨在解决大型音频语言模型（LALM）中普遍存在的“幻觉”问题（即生成与音频证据不符的内容）缺乏系统性评估工具的难题。为此，作者构建并发布了**HalluAudio**，这是首个大规模、多领域（语音、环境声、音乐）、多任务（二分类、多选、属性验证、开放生成）的人工验证音频幻觉检测基准，包含超过
论文解读
Qwen3.5-Omni Technical Report
这篇技术报告全面介绍了Qwen3.5-Omni，一个能够统一理解与生成文本、图像、音频和音视频内容的全模态大语言模型。**要解决的问题**是现有模型在实时交互、跨模态推理和自主智能体行为方面的局限性。**采用的方法**是基于“思考者-说话者”架构，引入了多项关键创新：1）思考者和说话者均采用混合注意