高效训练论文趋势 2023–2026

Scaling law、MoE、蒸馏与训练效率

统计范围：NeurIPS、ICLR、ICML、CVPR 四个会议，2023–2026 年，共 1,765 篇被归入该主题的论文。

主要结论

每年论文数

各会议论文数

按年份和会议统计

高效训练论文数，按年份和会议
年份NeurIPSICLRICMLCVPR合计
2023112678384346
2024207812280300
202519012611882516
2026177148169109603

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

全部高效训练论文

年份全部2023202420252026
会议全部NeurIPSICLRICMLCVPR

共 1,765 篇，第 5 / 36 页

  1. ICLR 2026MoL: Adaptive Mixture-of-Length Reasoning for Efficient Question Answering with Context
  2. ICLR 2026MOSS: Efficient and Accurate FP8 LLM Training with Microscaling and Automatic Scaling
  3. ICLR 2026NAIPv2: Debiased Pairwise Learning for Efficient Paper Quality Estimation
  4. ICLR 2026NeuCLIP: Efficient Large-Scale CLIP Training with Neural Normalizer Optimization
  5. ICLR 2026NLI : Non-uniform Linear Interpolation Approximation of Nonlinear Operations for Efficient LLMs Inference
  6. ICLR 2026No outlier channels but with outlier blocks
  7. ICLR 2026Not-a-Bandit: Provably No-Regret Drafter Selection in Speculative Decoding for LLMs
  8. ICLR 2026Optimal Aggregation of LLM and PRM Signals for Efficient Test-Time Scaling
  9. ICLR 2026Otters: An Energy-Efficient Spiking Transformer via Optical Time-to-First-Spike Encoding
  10. ICLR 2026Out of the Memory Barrier: A Highly Memory-Efficient Training System for LLMs with Million-Token Contexts
  11. ICLR 2026ParoQuant: Pairwise Rotation Quantization for Efficient Reasoning LLM Inference
  12. ICLR 2026PASER: Post-Training Data Selection for Efficient Pruned Large Language Model Recovery
  13. ICLR 2026PhaseFormer: From Patches to Phases for Efficient and Effective Time Series Forecasting
  14. ICLR 2026PiCa: Parameter-Efficient Fine-Tuning with Column Space Projection
  15. ICLR 2026Pre-training under infinite compute
  16. ICLR 2026ProxyAttn: Guided Sparse Attention via Representative Heads
  17. ICLR 2026QKV Projections Require a Fraction of Their Memory
  18. ICLR 2026Qronos: Correcting the Past by Shaping the Future... in Post-Training Quantization
  19. ICLR 2026Quantized Gradient Projection for Memory-Efficient Continual Learning
  20. ICLR 2026QuoKA: Query-Oriented KV Selection for Efficient LLM Prefill
  21. ICLR 2026RainPro-8: An Efficient Deep Learning Model to Estimate Rainfall Probabilities Over 8 Hours
  22. ICLR 2026Remaining-data-free Machine Unlearning by Suppressing Sample Contribution
  23. ICLR 2026Rethinking JEPA: Compute‑Efficient Video Self-Supervised Learning with Frozen Teachers
  24. ICLR 2026RMAAT: Astrocyte-Inspired Memory Compression and Replay for Efficient Long-Context Transformers
  25. ICLR 2026Scalable Random Wavelet Features: Efficient Non-Stationary Kernel Approximation with Convergence Guarantees
  26. ICLR 2026Scaling Laws Meet Model Architecture: Toward Inference-Efficient LLMs
  27. ICLR 2026Scaling Laws Revisited: Modeling the Role of Data Quality in Language Model Pretraining
  28. ICLR 2026Scaling Up, Speeding Up: A Benchmark of Speculative Decoding for Efficient LLM Test-Time Scaling
  29. ICLR 2026Scaling with Collapse: Efficient and Predictable Training of LLM Families
  30. ICLR 2026SecP-Tuning: Efficient Privacy-Preserving Prompt Tuning for Large Language Models via MPC
  31. ICLR 2026Sem-MoE: Semantic-aware Model-Data Collaborative Scheduling for Efficient MoE Inference
  32. ICLR 2026Semantic Voting: A Self-Evaluation-Free Approach for Efficient LLM Self-Improvement on Unverifiable Open-ended Tasks
  33. ICLR 2026SERE: Similarity-based Expert Re-routing for Efficient Batch Decoding in MoE Models
  34. ICLR 2026SparseEval: Efficient Evaluation of Large Language Models by Sparse Optimization
  35. ICLR 2026SPICE: Submodular Penalized Information–Conflict Selection for Efficient Large Language Model Training
  36. ICLR 2026SSDi8: Accurate and Efficient 8-bit Quantization for State Space Duality
  37. ICLR 2026STaMP: Sequence Transformation and Mixed Precision for Low-Precision Activation Quantization
  38. ICLR 2026Stopping Computation for Converged Tokens in Masked Diffusion-LM Decoding
  39. ICLR 2026Test-Time Efficient Pretrained Model Portfolios for Time Series Forecasting
  40. ICLR 2026TGM: A Modular and Efficient Library for Machine Learning on Temporal Graphs
  41. ICLR 2026TokenSeek: Memory Efficient Fine Tuning via Instance-Aware Token Ditching
  42. ICLR 2026Towards Greater Leverage: Scaling Laws for Efficient Mixture-of-Experts Language Models
  43. ICLR 2026Towards Lossless Memory-efficient Training of Spiking Neural Networks via Gradient Checkpointing and Spike Compression
  44. ICLR 2026ULD-Net: Enabling Ultra-Low-Degree Fully Polynomial Networks for Homomorphically Encrypted Inference
  45. ICLR 2026UltraMemV2: Memory Networks Scaling to 120B Parameters with Superior Long-Context Learning
  46. ICLR 2026UniCon: Unified Framework for Efficient Contrastive Alignment via Kernels
  47. ICLR 2026Unified and Efficient Multi-view Clustering from Probabilistic Perspective
  48. ICLR 2026Universal Model Routing for Efficient LLM Inference
  49. ICLR 2026Unlocking Full Efficiency of Token Filtering in Large Language Model Training
  50. ICLR 2026Unveiling Downstream Performance Scaling of LLMs: A Clustering-Based Perspective

在 Claude、Cursor 里直接问这个趋势

接上 Paperlist MCP 后，写综述、挑方向或准备投稿时，不用离开对话框就能拿到这页的数字：每年各会议的论文数、增长倍数、涨得最快的子方向和代表论文，AI 会按你的问题整理。

可以直接复制这句话问 AI

用 paperlist 查高效训练在 NeurIPS 2023–2026 的论文趋势，哪一年涨得最快？

AI 会调用的工具：get_topic_stats {"topic": "efficient-training"}

两分钟接入 Paperlist MCP（Claude Code、Cursor、Codex、ChatGPT）

常见问题

2023–2026 年四大会议一共有多少篇高效训练论文？
共 1,765 篇，按年份分别是 2023 年 346 篇、2024 年 300 篇、2025 年 516 篇、2026 年 603 篇。
NeurIPS 2026 有多少篇高效训练论文？
NeurIPS 2026 有 177 篇高效训练论文，上一届是 190 篇。
ICLR 2026 有多少篇高效训练论文？
ICLR 2026 有 148 篇高效训练论文，上一届是 126 篇。
ICML 2026 有多少篇高效训练论文？
ICML 2026 有 169 篇高效训练论文，上一届是 118 篇。
CVPR 2026 有多少篇高效训练论文？
CVPR 2026 有 109 篇高效训练论文，上一届是 82 篇。
哪个会议的高效训练论文最多？
NeurIPS，2023–2026 年共 499 篇。

「效率与系统」下的其他主题