Training-Inference Mismatch

Ask about this topic

Latest papers

CardsList

15 papers2 papers in the last four weeks, against 2 the four weeks before. 0.0% of all new papers.

  1. On the Origin and Evolution of Training-Inference Mismatch in LLM Reinforcement Learning

    Oct 29, 2026Yu Li, Yufa Zhou, Tian LanTraining-Inference MismatchRL for Language Models

    George Washington University · Duke University

  2. End-to-End MXFP8 Reinforcement Learning on Blackwell: What Matching Buys, When It Pays Off, and How It Fails

    Oct 26, 2026Wengee ZhangTraining-Inference MismatchRL Fine-Tuning

  3. Checkpoint Reload Explains Most of the Measured State Gap in Two State Discovery Reproductions

    Oct 26, 2026Weijia Han, Lisha QuTraining-Inference MismatchLatent World Models

    University of Washington

  4. Where Rounding Happens Matters: Rounding-Location Alignment Stabilizes LLM Reinforcement Learning

    Oct 23, 2026Yijing Ke, Tianle Cai, Ruosong WangRL for Language ModelsRL Stability

    ByteDance Seed · Peking University

  5. Parallel Time-Aligned Spiking Self-Attention for Consistent Integer-Valued Training and Spike-Driven Inference

    Oct 2, 2026Peng Xue, Wei Fang, Kaiwei Che +4Training-Inference MismatchSpiking Neural Networks

    Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, China · Pengcheng Laboratory, China · University of Chinese Academy of Sciences, China +2

  6. Rethinking Training-Inference Mismatch in LLM Reinforcement Learning: Where It Arises and How to Correct It

    Sep 26, 2026Tianrun Yu, Kaixiang Zhao, Shangzhe Li +4Training-Inference MismatchReinforcement Learning

    Brigham Young University · University of North Carolina at Chapel Hill

  7. Score Centering Stabilizes Off-policy Reinforcement Learning

    Sep 17, 2026Martin Marek, Max RyabininTraining-Inference MismatchRL for Language Models

    Together AI

  8. Reducing Pretraining-Generation Mismatch in Diffusion Language Models

    Aug 10, 2026Xiaocheng Lu, Huabin Liu, Song Guo +1Language Model PretrainingTraining-Inference Mismatch

    Inclusion AI · The Hong Kong University of Science and Technology

  9. Reformulate LLM Reinforcement Learning for Efficient Training under Black-box Discrepancy

    Jun 7, 2026Jiashun Liu, Runze Liu, Xu Wan +3Training-Inference MismatchRL for Language Models

    1Hong Kong University of Science and Technology · 2Zhejiang University · 3Tianjin University

  10. Double Preconditioning (DoPr): Optimization for Test-Time Performance, not Validation Loss

    Jun 4, 2026Thomas T. Zhang, Alok Shah, Yifei Zhang +3Deep Learning OptimizationTraining-Inference Mismatch

    University of Pennsylvania · Carnegie Mellon University · Amazon FAR

  11. Coupled Training with Privileged Information and Unlabeled Data

    May 22, 2026Jiahao Shi, Omar Hagrass, Jason M. KlusowskiLearning Using Privileged InformationSemi-Supervised Learning

    Department of Electrical and Computer Engineering, Princeton University · Department of Operations Research and Financial Engineering, Princeton University

  12. Diagnosing Training Inference Mismatch in LLM Reinforcement Learning via a Zero-Mismatch Reference

    May 14, 2026Tianle Zhong, Neiwen Ling, Yifan Pi +5Training-Inference MismatchReinforcement Learning

    ByteDance · The University of Virginia

  13. Compute Aligned Training: Optimizing for Test Time Inference

    Apr 27, 2026Adam Ousherovitch, Ambuj TewariTest-Time OptimizationInference-Time Scaling

    Department of Statistics University of Michigan

  14. Training-inference input alignment outweighs framework choice in longitudinal retinal image prediction

    Apr 18, 2026Liyin Chen, Nazlee Zebardast, Mengyu Wang +2Conditional Diffusion ModelsMedical Imaging

    Department of Ophthalmology, Massachusetts Eye and Ear, Harvard Medical School, Boston, MA, USA · Broad Institute of MIT and Harvard, Cambridge, MA, USA · Harvard Ophthalmology AI Lab, Schepens Eye Research Institute, Harvard Medical School, Boston, MA, USA +1

  15. Align Forward, Adapt Backward: Closing the Discretization Gap in Logic Gate Networks

    Mar 14, 2026Youngsung KimTraining-Inference MismatchNeural Network Optimization

    Department of Artificial Intelligence and Department of Electrical and Computer Engineering, Inha University, Incheon, Korea.