口语理解论文趋势 2023–2026

主要结论

• 口语理解论文从 2023 年的 40 篇增长到 2026 年的 61 篇，三年间扩大到约 1.5 倍。
• CVPR 是发口语理解论文最多的会议，2023–2026 年共 62 篇，占四个会议总量的 33%。
• 交叉方向里涨得最快的是多模态大模型：同时属于两个主题的论文从 2023 年的 3 篇增至 2026 年的 14 篇。
• 在「语音与音频」的 5 个主题里，它的 2026 年论文数排第 4；同组增长最快的音频生成是 4.3 倍，口语理解是 1.5 倍。

代表论文

按年份和会议均匀挑选，方便快速看到每一届的样貌

• NeurIPS 2026Denoise First, Orthogonalize Later: Understanding Momentum in Muon via Spectral Filtering
• ICML 2026A Single Layer to Explain Them All: Understanding Massive Values in Large Language Models
• ICLR 2026Neuron-Level Analysis of Cultural Understanding in Large Language Models
• CVPR 2026Consistent Instance Field for Dynamic Scene Understanding
• NeurIPS 2025The Graphon Limit Hypothesis: Understanding Neural Network Pruning via Infinite Width Analysis
• ICML 2025On Understanding Attention-Based In-Context Learning for Categorical Data
• ICLR 2025Toward Understanding In-context vs. In-weight Learning
• CVPR 2025HUSH: Holistic Panoramic 3D Scene Understanding using Spherical Harmonics
• NeurIPS 2024SD-Eval: A Benchmark Dataset for Spoken Dialogue Understanding Beyond Words
• ICML 2024Understanding Server-Assisted Federated Learning in the Presence of Incomplete Client Participation
• ICLR 2024Understanding Addition in Transformers
• CVPR 2024SurroundSDF: Implicit 3D Scene Understanding Based on Signed Distance Field
• NeurIPS 2023ATMAN: Understanding Transformer Predictions Through Memory Efficient Attention Manipulation
• ICML 2023How Do Transformers Learn Topic Structure: Towards a Mechanistic Understanding
• ICLR 2023Towards Understanding Why Mask Reconstruction Pretraining Helps in Downstream Tasks
• CVPR 2023AssemblyHands: Towards Egocentric Activity Understanding via 3D Hand Pose Estimation
• NeurIPS 2026EmbodiedObject: All-in-One Object Understanding with SLAT
• ICML 2026Understanding the Parameter Space Geometry of Transformers Encoding Boolean Functions
• ICLR 2026Towards Understanding the Shape of Representations in Protein Language Models
• CVPR 2026Lifting Unlabeled Internet-level Data for 3D Scene Understanding

全部口语理解论文

共 189 篇，第 1 / 4 页

1. CVPR 20264D-RGPT Toward Region-level 4D Understanding via Perceptual Distillation
2. CVPR 2026Blink Dynamic Visual Token Resolution for Enhanced Multimodal Understanding
3. CVPR 2026Consistent Instance Field for Dynamic Scene Understanding
4. CVPR 2026EarlyTom Early Token Compression Completes Fast Video Understanding
5. CVPR 2026EgoSound Benchmarking Sound Understanding in Egocentric Videos
6. CVPR 2026FPS-Bench A Benchmark for High Frame-Rate Video Understanding
7. CVPR 2026Latent Diffusion Inversion Requires Understanding the Latent Space
8. CVPR 2026Learning to Generate via Understanding Understanding-Driven Intrinsic Rewarding for Unified
9. CVPR 2026Lifting Unlabeled Internet-level Data for 3D Scene Understanding
10. CVPR 2026MA-Bench Towards Fine-grained Micro-Action Understanding
11. CVPR 2026Mistake Attribution Fine-Grained Mistake Understanding in Egocentric Videos
12. CVPR 2026MMLandmarks a Cross-View Instance-Level Benchmark for Geo-Spatial Understanding
13. CVPR 2026RISE Single Static Radar-based Indoor Scene Understanding
14. CVPR 2026TimeRipples Accelerating vDiTs by Understanding the Spatio-Temporal Correlations in Latent
15. CVPR 2026Towards Generalized Representations for Low-Light Understanding When Signal Constancy Meets
16. CVPR 2026ViKey Enhancing Temporal Understanding in Videos via Visual Prompting
17. CVPR 2026When Token Pruning is Worse than Random Understanding Visual Token
18. CVPR 2026World in a Frame Understanding Culture Mixing as a New
19. ICLR 2026MMSU: A Massive Multi-task Spoken Language Understanding and Reasoning Benchmark
20. ICLR 2026Neuron-Level Analysis of Cultural Understanding in Large Language Models
21. ICLR 2026Towards Understanding the Shape of Representations in Protein Language Models
22. ICLR 2026Understanding In-Context Learning on Structured Manifolds: Bridging Attention to Kernel Methods
23. ICLR 2026Understanding Language Prior of LVLMs by Contrasting Chain-of-Embedding
24. ICLR 2026Understanding the Mixture-of-Experts with Nadaraya-Watson Kernel
25. ICML 20263DPoV: Improving 3D understanding via Patch Ordering on Videos
26. ICML 2026A Single Layer to Explain Them All: Understanding Massive Values in Large Language Models
27. ICML 2026AudioChat: Unified Audio Storytelling, Editing, and Understanding with Transfusion Forcing
28. ICML 2026CURVE: Learning Causality-Inspired Invariant Representations for Robust Scene Understanding via Uncertainty-Guided Regularization
29. ICML 2026From Denoising to De-Channeling: Integrating Physical Channel Priors into Diffusion Models for Radio Signal Understanding
30. ICML 2026MetaphorVU: Towards Metaphorical Video Understanding
31. ICML 2026Positive Distribution Shift as a Framework for Understanding Tractable Learning
32. ICML 2026Toward Understanding Adversarial Distillation: Why Robust Teachers Fail
33. ICML 2026Towards Understanding Steering Strength
34. ICML 2026Understanding Dynamics of Adam in Zero-Sum Games: An ODE Approach
35. ICML 2026Understanding SAM through Minimax Perspective
36. ICML 2026Understanding the Parameter Space Geometry of Transformers Encoding Boolean Functions
37. NeurIPS 2026Beyond Oversquashing: Understanding Signal Propagation in GNNs Via Observables
38. NeurIPS 2026Beyond Single Expert: Harmonizing Diverse Visual Priors in MLLMs for Spatial Understanding
39. NeurIPS 2026Denoise First, Orthogonalize Later: Understanding Momentum in Muon via Spectral Filtering
40. NeurIPS 2026EmbodiedObject: All-in-One Object Understanding with SLAT
41. NeurIPS 2026Emergent Semantic Role Understanding in Language Models
42. NeurIPS 2026Evaluating Multimodal Narrative Understanding of Popular Hollywood Films
43. NeurIPS 2026Incorporating Hierarchical Semantics in Sparse Autoencoder Architectures
44. NeurIPS 2026Overcoming State Inertia in Full-Duplex Spoken Language Models via Activation Steering
45. NeurIPS 2026RAVEN-Bench: A Paired EO-IR Video QA Benchmark for Aerial Multimodal Understanding
46. NeurIPS 2026SceneScaffold: Active Scene-State Construction for Unified 3D Scene Understanding
47. NeurIPS 2026Shifting the Gradient: Understanding How Defensive Training Methods Protect Language Model Integrity
48. NeurIPS 2026StakeBench: Evaluating Language Understanding Grounded in Market Commitment
49. NeurIPS 2026TELEVAL: A Benchmark Designed for Spoken Language Models in Chinese Interactive Scenarios
50. NeurIPS 2026U-Bench: A Comprehensive Understanding of U-Net through 100-Variant Benchmarking

常见问题

2023–2026 年四大会议一共有多少篇口语理解论文？

共 189 篇，按年份分别是 2023 年 40 篇、2024 年 38 篇、2025 年 50 篇、2026 年 61 篇。

NeurIPS 2026 有多少篇口语理解论文？

NeurIPS 2026 有 25 篇口语理解论文，上一届是 14 篇。

ICLR 2026 有多少篇口语理解论文？

ICLR 2026 有 6 篇口语理解论文，上一届是 7 篇。

ICML 2026 有多少篇口语理解论文？

ICML 2026 有 12 篇口语理解论文，上一届是 10 篇。

CVPR 2026 有多少篇口语理解论文？

CVPR 2026 有 18 篇口语理解论文，上一届是 19 篇。

哪个会议的口语理解论文最多？

CVPR，2023–2026 年共 62 篇。