推理加速论文趋势 2023–2026

主要结论

• 推理加速论文从 2023 年的 54 篇增长到 2026 年的 91 篇，三年间扩大到约 1.7 倍。
• NeurIPS 是发推理加速论文最多的会议，2023–2026 年共 90 篇，占四个会议总量的 35%。
• 交叉方向里涨得最快的是深度学习 · large language models：同时属于两个主题的论文从 2023 年的 1 篇增至 2026 年的 9 篇。
• 在「效率与系统」的 5 个主题里，它的 2026 年论文数排第 5；同组增长最快的硬件协同是 4.4 倍，推理加速是 1.7 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 2026A Method-Class Divide in Sub-4-Bit Quantization:\\Iterative vs.\ Single-Pass Sensitivity to Data Composition
• ICML 2026Consistency Deep Equilibrium Models
• ICLR 2026OrderDP: A Theoretically Guaranteed Lossless Dynamic Data Pruning Framework
• CVPR 2026GeoRK2 Geometry-Guided Runge-Kutta Integration for Diffusion Transformer Acceleration
• NeurIPS 2025PolarQuant: Leveraging Polar Transformation for Key Cache Quantization and Decoding Acceleration
• ICML 2025RePaViT: Scalable Vision Transformer Acceleration via Structural Reparameterization on Feedforward Network Layers
• ICLR 2025Progressive distillation induces an implicit curriculum
• CVPR 2025Adapting Dense Matching for Homography Estimation with Grid-based Acceleration
• NeurIPS 2024Provable Acceleration of Nesterov's Accelerated Gradient for Asymmetric Matrix Factorization and Linear Neural Networks
• ICML 2024Diversified Batch Selection for Training Acceleration
• ICLR 2024Embarrassingly Simple Dataset Distillation
• CVPR 2024Data-Free Quantization via Pseudo-label Filtering
• NeurIPS 2023Continuous-time Analysis of Anchor Acceleration
• ICML 2023Auxiliary Modality Learning with Generalized Curriculum Distillation
• ICLR 2023Accurate Neural Training with 4-bit Matrix Multiplications at Standard Formats
• CVPR 2023NIPQ: Noise Proxy-Based Integrated Pseudo-Quantization
• NeurIPS 2026AcceleGrad#: Adaptive Geometry-Aware Acceleration
• ICML 2026PsumQuant: In-line Post-training Partial Sum Quantizer for Energy Efficient NPU Inference
• ICLR 2026NLI : Non-uniform Linear Interpolation Approximation of Nonlinear Operations for Efficient LLMs Inference
• CVPR 2026SODA Sensitivity-Oriented Dynamic Acceleration for Diffusion Transformer

全部推理加速论文

共 68 篇，第 1 / 2 页

1. ICML 2026ASTRA: Communication-Efficient Acceleration for Multi-Device Transformer Inference
2. ICML 2026Attn-QAT: 4-Bit Attention With Quantization-Aware Training
3. ICML 2026Budget-Constrained Step-Level Diffusion Caching
4. ICML 2026CoCoQuant: Breaking the Bandwidth Wall via Co-Optimized Communication and Computation Quantization
5. ICML 2026Consistency Deep Equilibrium Models
6. ICML 2026DC-Leap: Training-Free Acceleration of dLLMs via Draft-Guided Contiguous Leaping Decoding
7. ICML 2026FAFO: Lossy KV Cache Compression for Lossless Inference Acceleration via Draftless Fumble Decoding
8. ICML 2026FasterVAR: Plug-and-Play Acceleration for Visual Autoregressive Models
9. ICML 2026Gated Relational Alignment via Confidence-based Distillation for Efficient VLMs
10. ICML 2026Improving Classifier-Free Guidance of Flow Matching via Manifold Projection
11. ICML 2026Inference of Online Newton Methods with Nesterov's Accelerated Sketching
12. ICML 2026LearniBridge: Learnable Calibration of Feature Caching for Diffusion Models Acceleration
13. ICML 2026Lie-Algebraic Acceleration of Neural Koopman Dynamics
14. ICML 2026MAnchors: Memorization-Based Acceleration of Anchors via Rule Reuse and Transformation
15. ICML 2026MAPS: Memory-Aware Predictive Scheduling Framework for Large Language Model Serving
16. ICML 2026Matrix-Free GPU Semidefinite Programming for Quantum Ordered Search at the k=6 Frontier
17. ICML 2026PsumQuant: In-line Post-training Partial Sum Quantizer for Energy Efficient NPU Inference
18. ICML 2026Pushing the Limits of Block Rotations in Post-Training Quantization
19. ICML 2026RSTR: Reducing SpatioTemporal Redundancy in Diffusion Transformers
20. ICML 2026RT-Lynx: Putting the GEMM Sparsity In a Right Way for Diffusion Models
21. ICML 2026Semantic Cache Distillation: Efficient State Transfer via Reuse and Selective Patching
22. ICML 2026SparseInfer: Accelerating Large Language Model Inference with Semantics-Inspired Adaptive Sparse Activation
23. ICML 2026UniSVQ: 2-bit Unified Scalar-Vector Quantization
24. ICML 2026WaterSIC: Information-Theoretically (Near) Optimal \\Linear Layer Quantization
25. ICML 2026WAVE: Window-Aware Vocabulary-Efficient Early-Exit for Training-Free LLM Acceleration
26. ICML 2026XSpecMesh: Quality-Preserving Auto-Regressive Mesh Generation Acceleration via Multi-Head Speculative Decoding
27. ICML 2025any4: Learned 4-bit Numeric Representation for LLMs
28. ICML 2025BanditSpec: Adaptive Speculative Decoding via Bandit Algorithms
29. ICML 2025COGNATE: Acceleration of Sparse Tensor Programs on Emerging Hardware using Transfer Learning
30. ICML 2025Compress then Serve: Serving Thousands of LoRA Adapters with Little Overhead
31. ICML 2025Constant Stepsize Local GD for Logistic Regression: Acceleration by Instability
32. ICML 2025Distilling the Knowledge in Data Pruning
33. ICML 2025Ensemble Distribution Distillation via Flow Matching
34. ICML 2025Fast Large Language Model Collaborative Decoding via Speculation
35. ICML 2025Learning to Keep a Promise: Scaling Language Model Decoding Parallelism with Learned Asynchronous Decoding
36. ICML 2025Matryoshka Quantization
37. ICML 2025Metadata Conditioning Accelerates Language Model Pre-training
38. ICML 2025On Exact Bit-level Reversible Transformers Without Changing Architecture
39. ICML 2025Optimizing Large Language Model Training Using FP4 Quantization
40. ICML 2025Quamba2: A Robust and Scalable Post-training Quantization Framework for Selective State Space Models
41. ICML 2025RePaViT: Scalable Vision Transformer Acceleration via Structural Reparameterization on Feedforward Network Layers
42. ICML 2025SADA: Stability-guided Adaptive Diffusion Acceleration
43. ICML 2025SageAttention2: Efficient Attention with Thorough Outlier Smoothing and Per-thread INT4 Quantization
44. ICML 2025Scalable Model Merging with Progressive Layer-wise Distillation
45. ICML 2025SKIM: Any-bit Quantization Pushing The Limits of Post-Training Quantization
46. ICML 2025Sparsing Law: Towards Large Language Models with Greater Activation Sparsity
47. ICML 2025Streamline Without Sacrifice - Squeeze out Computation Redundancy in LMM
48. ICML 2025When Dynamic Data Selection Meets Data Augmentation: Achieving Enhanced Training Acceleration
49. ICML 2024A2Q+: Improving Accumulator-Aware Weight Quantization
50. ICML 2024Accelerating Transformer Pre-training with 2:4 Sparsity

常见问题

2023–2026 年四大会议一共有多少篇推理加速论文？

共 259 篇，按年份分别是 2023 年 54 篇、2024 年 50 篇、2025 年 64 篇、2026 年 91 篇。

NeurIPS 2026 有多少篇推理加速论文？

NeurIPS 2026 有 33 篇推理加速论文，上一届是 15 篇。

ICLR 2026 有多少篇推理加速论文？

ICLR 2026 有 17 篇推理加速论文，上一届是 18 篇。

ICML 2026 有多少篇推理加速论文？

ICML 2026 有 26 篇推理加速论文，上一届是 22 篇。

CVPR 2026 有多少篇推理加速论文？

CVPR 2026 有 15 篇推理加速论文，上一届是 9 篇。

哪个会议的推理加速论文最多？

NeurIPS，2023–2026 年共 90 篇。