#** #Video #LLMs #Token #Interface #Alignment #多模态整合 #流形对齐 #单模态数据

✅ 7.5/10 | 前25% | #** | #Video | #LLMs #Token | arxiv

← 返回 2026-05-23 语音/音乐/音频论文速递

共分析 123 篇论文

⚡ 今日概览

📥 抓取 123 篇 → 🔬 深度分析完成

🏷️ 热门方向

📊 论文评分排行榜（123 篇，按分数降序）

📋 论文列表

🥇 INFER: Learning Implicit Neural Frequency Response Fields for Confined Acoustic Environments

🔥 8.5/10 | 前25% | arxiv

🥈 VocSim A Training-free Benchmark for Zero-shot Content Identity in Single-source Audio

🔥 8.3/10 | 前25% | arxiv

🥉 CMI-RewardBench: Evaluating Music Reward Models with Compositional Multimodal Instruction

✅ 7.2/10 | 前50% | arxiv

4. Language Model Augmented Semi-Supervised Statistical Inference

🔥 8.2/10 | 前25% | arxiv

5. DiscoForcing: A Unified Framework for Real-Time Audio-Driven Character Control with Diffusion Forcing

🔥 8.2/10 | 前25% | arxiv

6. Abstraction Induces the Brain Alignment of Language and Speech Models

🔥 8.0/10 | 前25% | arxiv

7. Alethia: a Foundational Encoder for Voice Deepfakes

✅ 7.5/10 | 前25% | arxiv

8. OmniDenseCap: Scripting Multi-Scene Videos with Time-Aware and Structural Audio-Visual Captions

🔥 8.0/10 | 前25% | arxiv

9. FoeGlass: When Simple In-Context Learning Is Enough for Red Teaming Audio Deepfake Detectors

🔥 8.0/10 | 前25% | arxiv

10. E-VAds: An E-commerce Short Videos Understanding Benchmark for MLLMs

🔥 8.0/10 | 前25% | arxiv

11. BEAT: Tokenizing and Generating Symbolic Music by Uniform Temporal Steps

🔥 8.0/10 | 前25% | arxiv

12. Pianist Transformer: Towards Expressive Piano Performance Rendering via Scalable Self-Supervised Pre-Training

✅ 7.8/10 | 前25% | arxiv

13. DreamID-Omni: Unified Framework for Controllable Human-Centric Audio-Video Generation

✅ 7.8/10 | 前25% | arxiv

14. Real-World Unsupervised Models Generalize to Predict Brain Responses to Out-of-Distribution Stimuli

✅ 7.8/10 | 前25% | arxiv

15. AudioMosaic: Contrastive Masked Audio Representation Learning

✅ 7.5/10 | 前25% | arxiv

16. Self-Guidance: Enhancing Neural Codecs via Decoder Manifold Alignment

✅ 7.5/10 | 前25% | arxiv

17. LynX: Token Interface Alignment for Video+X LLMs

✅ 7.5/10 | 前25% | arxiv

18. Spherical Procrustes Alignment for Reliable Medical Audio Diagnosis

✅ 7.5/10 | 前25% | arxiv

19. MoST: Mixing Speech and Text with Modality-Aware Mixture of Experts

✅ 7.5/10 | 前25% | arxiv

20. Self-Supervised Flow Matching for Scalable Multi-Modal Synthesis

✅ 7.5/10 | 前25% | arxiv

21. LightAVSeg: Lightweight Audio-Visual Segmentation

✅ 7.5/10 | 前25% | arxiv

22. Robust Signal Enhancement via Fractional Detail Views and Knowledge Guided Multi-view Fusion

✅ 7.5/10 | 前25% | arxiv

23. EchoingPixels: Aliasing-Resistant Joint Token Reduction for Audio-Visual LLMs

✅ 7.5/10 | 前25% | arxiv

24. Long Grounded Thoughts: Synthesizing Grounded Visual Problems and Distilling Reasoning Chains at Scale

✅ 7.5/10 | 前25% | arxiv

25. OmniVideo-R1: Reinforcing Audio-visual Reasoning with Query Intention and Modality Attention

✅ 7.5/10 | 前25% | arxiv

26. Ariadne’s Thread of LipSync: Unraveling Forgeries via Inconsistency between Lip Motions and Head Poses

✅ 7.5/10 | 前25% | arxiv

27. AVI-Bench: Toward Human-like Audio-Visual Intelligence of Omni-MLLMs

✅ 7.5/10 | 前25% | arxiv

28. Simultaneous Speech-to-Speech Translation Without Aligned Data

🔥 8.2/10 | 前25% | arxiv

29. PhoStream: Benchmarking Real-World Streaming for Omnimodal Assistants in Mobile Scenarios

✅ 7.5/10 | 前25% | arxiv

30. OmniSIFT: Modality-Asymmetric Token Compression for Efficient Omni-modal Large Language Models

✅ 7.5/10 | 前25% | arxiv

31. Speech-Audio Compositional Attacks on Multimodal LLMs and Their Defense with SALMONN-Guard

✅ 7.5/10 | 前25% | arxiv

32. Convex Low-resource Accent-Robust Language Detection in Speech Recognition

✅ 7.5/10 | 前25% | arxiv

33. PhaseCoder: Microphone Geometry-Agnostic Spatial Audio Understanding for Multimodal LLMs

✅ 7.5/10 | 前25% | arxiv

34. Listening Through the Noise: Cauchy-Driven Diffusion Bridges for Robust Gastrointestinal Auscultation and Clinical Benchmarking

✅ 7.5/10 | 前25% | arxiv

35. Dual-View Predictive Diffusion: Lightweight Speech Enhancement via Spectrogram-Image Synergy

✅ 7.5/10 | 前25% | arxiv

36. Stream RAG: Instant and Accurate Spoken Dialogue Systems with Streaming Tool Usage

✅ 7.5/10 | 前25% | arxiv

37. NAACA: Training-Free NeuroAuditory Attentive Cognitive Architecture with Oscillatory Working Memory for Salience-Driven Attention Gating

✅ 7.5/10 | 前25% | arxiv

38. MedMosaic: A Challenging Large Scale Benchmark of Diverse Medical Audio

✅ 7.5/10 | 前25% | arxiv

39. Verifiable Multimodal Reasoning: Fact-level Attribution with Multimodal Sources

✅ 7.5/10 | 前25% | arxiv

40. MusicDET: Zero-Shot AI-Generated Music Detection

✅ 7.5/10 | 前25% | arxiv

41. PCRNet: Phase-aware Complex Refinement Network for EEG-based Auditory Attention Decoding

✅ 7.5/10 | 前25% | arxiv

42. SARSteer: Safeguarding Large Audio Language Models via Safe-Ablated Refusal Steering

✅ 7.5/10 | 前25% | arxiv

43. STAR-VAE: Structured Topology-Aware Regularization for Audio Reconstruction and Generation

✅ 7.5/10 | 前25% | arxiv

44. Hidden in Plain Tokens: Simply Robust, Gradient-Free Watermark for Synthetic Audio

✅ 7.5/10 | 前25% | arxiv

45. AVGen-Bench: A Task-Driven Benchmark for Multi-Granular Evaluation of Text-to-Audio-Video Generation

✅ 7.3/10 | 前50% | arxiv

46. Bridging the Stability-Expressivity Gap: Synthetic Data Scaling and Preference Alignment for Low-Resource Spoken Language Models

✅ 7.3/10 | 前50% | arxiv

47. AVTrack: Audio-Visual Speaker Tracking in Complex Scenes

✅ 7.3/10 | 前50% | arxiv

48. Bioacoustic Geolocation: Species Sounds as Geographic Signals

✅ 7.2/10 | 前50% | arxiv

49. ADEPT: RL-Aligned Agentic Decoding of Emotion via Evidence Probing Tools — From Consensus Learning to Ambiguity-Driven Emotion Reasoning

✅ 7.2/10 | 前50% | arxiv

50. MECAT: A Multi-Experts Constructed Benchmark for Fine-Grained Audio Understanding Tasks

✅ 7.2/10 | 前50% | arxiv

51. SPEAR: A Unified SSL Framework for Learning Speech and Audio Representations

✅ 7.2/10 | 前50% | arxiv

52. PADS-TAL: Padding-Annealed Diffusion Sampling in Text-Aware Latent Space for Robust and Diverse Text-to-Music Generation

✅ 7.2/10 | 前50% | arxiv

53. Multimodal Latent Language Modeling with Next-Token Diffusion

✅ 7.2/10 | 前50% | arxiv

54. Query-Based Asymmetric Modeling with Decoupled Input–Output Rates for Speech Restoration

✅ 7.0/10 | 前50% | arxiv

55. AgentSteerTTS: A Multi-Agent Closed-Loop Framework for Composite-Instruction Text-to-Speech

✅ 7.0/10 | 前50% | arxiv

56. Optimality of FSQ tokens for continuous diffusion for categorical data with application to text-to-speech

✅ 7.0/10 | 前50% | arxiv

57. JAEGER: Joint 3D Audio-Visual Grounding and Reasoning in Simulated Physical Environments

✅ 7.0/10 | 前50% | arxiv

58. SonicMaster: Towards Controllable All-in-One Music Restoration and Mastering

✅ 7.0/10 | 前50% | arxiv

59. VIBE: Disentangling Social Dynamics via Kinematics-Informed Variational Inference for Behavioral Emotion

✅ 7.0/10 | 前50% | arxiv

60. Reasoning LLM Improves Speaker Recognition in Long-form TV Dramas

✅ 7.0/10 | 前50% | arxiv

61. A Semantically Consistent Dataset for Data-Efficient Query-Based Universal Sound Separation

✅ 7.0/10 | 前50% | arxiv

62. The Silent Thought: Modeling Internal Cognition in Full-Duplex Spoken Dialogue Models via Latent Reasoning

🔥 8.5/10 | 前25% | arxiv

63. Learning Tight Rejection Boundaries without Negatives for Strict One-Class Audio Deepfake Detection

✅ 7.0/10 | 前50% | arxiv

64. Quaternion Self-Attention with Shared Scores

✅ 7.0/10 | 前50% | arxiv

65. Bridging Your Imagination with Audio-Video Generation via a Unified Director

✅ 7.0/10 | 前50% | arxiv

66. TextME: Bridging Unseen Modalities Through Text Descriptions

✅ 7.0/10 | 前50% | arxiv

67. ReGen: Hierarchical Multi-Prompt Representation Generation for Efficient Waveform Diffusion Models

✅ 7.0/10 | 前50% | arxiv

68. Polyphonia: Training-Free Context-Aware Music Editing with Acoustic-Informed Attention Calibration

✅ 7.0/10 | 前50% | arxiv

69. TMD-Bench: A Multi-Level Evaluation Paradigm for Music–Dance Co-Generation

✅ 7.0/10 | 前50% | arxiv

70. Omni-Perception Policy Optimization for Multimodal Emotion Reasoning

✅ 7.0/10 | 前50% | arxiv

71. Acoustic Interference: A New Paradigm Weaponizing Acoustic Latent Semantic for Universal Jailbreak against Large Audio Language Models

✅ 7.0/10 | 前50% | arxiv

72. AudioChat: Unified Audio Storytelling, Editing, and Understanding with Transfusion Forcing

✅ 7.0/10 | 前50% | arxiv

73. Do Audio LLMs Listen or Read? Analyzing and Mitigating Paralinguistic Failures with VoxParadox

✅ 6.9/10 | 前50% | arxiv

74. From Talking to Singing: A New Challenge for Audio-Visual Deepfake Detection

✅ 6.8/10 | 前50% | arxiv

75. Multiple Choice Learning of Low-Rank Adapters for Language Modeling

✅ 6.8/10 | 前50% | arxiv

76. Multimodal Fusion via Self-Consistent Task-Gradient Fields

✅ 6.8/10 | 前50% | arxiv

77. Position: Beyond Text The Text-Centric Bias in Foundation Models Must Be Revisited for a Speech-First Future

✅ 6.8/10 | 前50% | arxiv

78. MetaBio: Learning from metadata for bioacoustics foundation models

✅ 6.5/10 | 前50% | arxiv

79. Any-Diffusion: Unified Multimodal Understanding and Generation with Masked Discrete Diffusion

✅ 6.5/10 | 前50% | arxiv

80. SAM Audio: Segment Anything in Audio

✅ 6.5/10 | 前50% | arxiv

81. CoCoEmo: Composable and Controllable Human-Like Emotional TTS via Activation Steering

✅ 6.5/10 | 前50% | arxiv

82. HyperPotter: Spell the Charm of High-Order Interactions in Audio Deepfake Detection

✅ 6.5/10 | 前50% | arxiv

83. Joint Enhancement and Classification using Coupled Diffusion Models of Signals and Logits

✅ 6.5/10 | 前50% | arxiv

84. Hearing Without Noticing? Attention-Aware Stealthy Black-box Adversarial Audio Attacks

✅ 6.5/10 | 前50% | arxiv

85. Two-dimensional quantization for geometry-aware audio coding

✅ 6.5/10 | 前50% | arxiv

86. SALSA-V: Shortcut-Augmented Long-form Synchronized Audio from Videos

arxiv

87. REST: Diffusion-based Real-time End-to-end Streaming Talking Head Generation via ID-Context Caching and Asynchronous Streaming Distillation

✅ 6.5/10 | 前50% | arxiv

88. AuTAgent: A Reinforcement Learning Framework for Tool-Augmented Audio Reasoning

✅ 6.5/10 | 前50% | arxiv

89. Characterizing the Predictive Impact of Modalities with Supervised Latent-Variable Modeling

✅ 6.5/10 | 前50% | arxiv

90. Group Cognition Learning: Making Everything Better Through Controlled Two-Stage Agents Collaboration

✅ 6.5/10 | 前50% | arxiv

91. Rethinking Attention in Spiking Transformers: Overcoming Density Bias with Set Similarity

✅ 6.5/10 | 前50% | arxiv

92. T2AV-Compass: Towards Unified Evaluation for Text-to-Audio-Video Generation

✅ 6.5/10 | 前50% | arxiv

93. S3Audio: Towards Streaming Synchronized Spatial Audio Generation via Autoregressive Diffusion Transformer

✅ 6.5/10 | 前50% | arxiv

94. Sparse Autoencoders for Interpretable Emotion Control in Text-to-Speech

✅ 6.5/10 | 前50% | arxiv

95. BAT: Better Audio Transformer Guided by Convex Gated Probing

✅ 6.5/10 | 前50% | arxiv

96. AG-REPA: Causal Layer Selection for Representation Alignment in Audio Flow Matching

✅ 6.5/10 | 前50% | arxiv

97. CoLA: Cross-Modal Low-rank Adaptation for Multimodal Downstream Tasks

✅ 6.5/10 | 前50% | arxiv

98. Neural-Inspired Modeling of Auditory Selection and Compensation for Audio-Visual Speech Separation

✅ 6.5/10 | 前50% | arxiv

99. FutureOmni: Evaluating Future Forecasting from Omni-Modal Context for Multimodal LLMs

✅ 6.5/10 | 前50% | arxiv

100. ProactiveLLM: Learning Active Interaction for Streaming Large Language Models

✅ 6.0/10 | 前50% | arxiv

101. video-SALMONN S: Memory-Enhanced Streaming Audio-Visual LLM

✅ 6.0/10 | 前50% | arxiv

102. Zero-Shot Rankability: Revealing Latent Ordinal Structure in Multimodal Large Language Models via Language

✅ 6.0/10 | 前50% | arxiv

103. Scaling Transformers for End-to-End Discrete Audio Tokenization

✅ 6.0/10 | 前50% | arxiv

104. Evaluating and Rewarding LALMs for Expressive Role-Play TTS via Mean Continuation Log-Probability

✅ 6.5/10 | 前50% | arxiv

105. Unlocking Speech–Text Compositional Powers: Instruction-Following Speech Language Models without Instruction Tuning

📝 5.8/10 | 前50% | arxiv

106. Probing Cross-modal Information Hubs in Audio-Visual LLMs

📝 5.5/10 | 前50% | arxiv

107. OmniShow: Orchestrating Multimodal Conditions for Human-Object Interaction Video Generation

📝 5.5/10 | 前50% | arxiv

108. Sparse Tokens Suffice: Jailbreaking Audio Language Models via Token-Aware Gradient Optimization

✅ 6.0/10 | 前50% | arxiv

109. PHALAR: Phasors for Learned Musical Audio Representations

📝 5.5/10 | 前50% | arxiv

110. Scaling Laws in Model Fine-tuning for Audio DeepFake Detection

📝 5.0/10 | 后50% | arxiv

111. PRIM：Cooperative Dynamic Token Compression for Efficient Large Multimodal Models

📝 4.8/10 | 后50% | arxiv

112. Towards Understanding Modality Interaction in Multimodal Language Models via Partial Information Decomposition

📝 4.5/10 | 后50% | arxiv

113. From Inpainting to Editing: Unlocking Robust Mask-Free Visual Dubbing via Generative Bootstrapping

📝 4.3/10 | 后50% | arxiv

114. SONAR: Spectral‑Contrastive Audio Residuals for Generalizable Deepfake Detection

📝 4.0/10 | 后50% | arxiv

115. MoshiRAG: Asynchronous Knowledge Retrieval for Full-Duplex Speech Language Models

📝 3.8/10 | 后50% | arxiv

116. STARCaster: Spatio-Temporal AutoRegressive Video Diffusion for Identity- and View-Aware Talking Portraits

📝 3.5/10 | 后50% | arxiv

117. WaveSSM: Multiscale State-Space Models for Non-stationary Signal Attention

📝 3.5/10 | 后50% | arxiv

118. \(\tau\)-Voice: Benchmarking Full-Duplex Voice Agents on Real-World Domains

📝 3.5/10 | 后50% | arxiv

119. FakeWorld 1.0: An Omni modal Benchmark for Fake Media and Content

📝 3.5/10 | 后50% | arxiv

120. LALM-as-a-Judge: Benchmarking Large Audio-Language Models for Safety Evaluation in Multi-Turn Spoken Dialogues

📝 3.5/10 | 后50% | arxiv

121. IVQ: Structured and Lightweight Vector Quantization via Binary Hierarchical Composition Inspired by \(\textit{IChing}\)

📝 3.2/10 | 后50% | arxiv

122. MFCL Audio: An Audio Function Calling Evaluation for Large Language Models

📝 3.0/10 | 后50% | arxiv

123. Position: Towards Responsible Evaluation for Text-to-Speech

📝 2.6/10 | 后50% | arxiv